Palaeomagnetic secular variation recorded by lavas from the Taupo Volcanic Zone, New Zealand

In order to understand the origin, temporal behaviour and spatial characteristics of Earth’s magnetic field, globally distributed records of the palaeomagnetic direction and absolute palaeointensity are required. However a paucity of data from the southern hemisphere significantly limits the resolution of global field models, particularly on short time-scales. In this thesis new, high quality palaeomagnetic data from volcanic materials sampled within the Taupo Volcanic Zone, New Zealand are presented, with a focus on the Tongariro and Okataina Volcanic Centre. New palaeomagnetic directions were obtained from 19 andesitic or rhyolitic lavas, of which 10 also produced successful palaeointensity results. Palaeointensity experiments were conducted using a combination of traditional Thellier-type thermal, and microwave techniques. Detailed magneto-mineralogical investigations carried out alongside these experiments helped to characterise the primary remanence carriers and to justify the reliability of the results. The study also revises the age controls and results from earlier palaeomagnetic studies on Holocene volcanic materials from the area. All new or revised data are summarized into a new data compilation for New Zealand, which includes 24 directions and ten palaeointensities dated between 1886 AD and 15,000 yrs BP. The new directional data reproduces the features of the most recently published continuous record from Lake Mavora (Fiordland, New Zealand), although with directions ranging in their extremes from 321° (west) to 26° (east) declination and -82 to -49° in inclination, the discrete dataset describes somewhat larger amplitude swings.

In this thesis new, high quality palaeomagnetic data from volcanic materials sampled within the Taupo Volcanic Zone, New Zealand are presented, with a focus on the Tongariro and Okataina Volcanic Centre.
New palaeomagnetic directions were obtained from 19 andesitic or rhyolitic lavas, of which 10 also produced successful palaeointensity results. Palaeointensity experiments were conducted using a combination of traditional Thellier-type thermal, and microwave techniques. Detailed magneto-mineralogical investigations carried out alongside these experiments helped to characterise the primary remanence carriers and to justify the reliability of the results.
The study also revises the age controls and results from earlier palaeomagnetic studies on Holocene volcanic materials from the area. All new or revised data are summarized into a new data compilation for New Zealand, which includes 24 directions and ten palaeointensities dated between 1886 AD and 15,000 yrs BP.
The new directional data reproduces the features of the most recently published continuous record from Lake Mavora (Fiordland, New Zealand), although with directions ranging in their extremes from 321° (west) to 26° (east) declination and -82 to -49° in inclination, the discrete dataset describes somewhat larger amplitude swings.
i ii With few exceptions, the new palaeointensity dataset describes a steady increase in the palaeointensity throughout the Holocene, from 37.0 ± 5.7 μT obtained from a pre-8 ka lava to 70.6 ± 4.1 μT from the youngest (≤ 500 yrs BP) flows sampled. A similar trend is also predicted by the latest global field model pfm9k. Furthermore, the data falls within the range of palaeointensity variation suggested by the Mavora record. The dataset roughly agrees with a global VADM reconstruction in the early Holocene (> 5000 yrs BP), but yields values significantly above the global trend in the late Holocene (< 1000 yrs BP) which supports the presence of significant non-dipolar components over the SW Pacific region in the time period, visible in global field models and from continuous PSV records.
A comparison of the directional records with the Mavora Curve provided refinement of age estimates of five lava flows from the Tongariro Volcanic Centre, from uncertainties in the range of 2-3000 years. The new palaeomagnetic emplacement age estimates for these flows have age brackets as short as 500 years and thus highlight different phases of the young cone building eruptive activity on Ruapehu volcano.
x                        Tables   Table 3-

Motivation
In contrast to all other terrestrial planets within the solar system, the Earth generates a largely dipolar and strong magnetic field. This effectively shields us and our planet's atmosphere and surface from harmful solar radiation, a stream of charged particles emitted from the sun. However, the Earth's magnetic field is subject to changes throughout time. Periodic field fluctuations ranging from years to millennia are called secular variation or palaeosecular variation (PSV) when referring to changes that predate direct magnetic observations. The most extreme phenomenon experienced by the Earth's magnetic field is the magnetic polarity reversal, which occurs in undefined intervals between thousands and millions of years. Shorter lived events that involve a sudden decrease in the magnetic field intensity and rapid movement of the poles to lower latitudes are called geomagnetic excursions -the last excursion occurred 33,000 years ago (Roberts, 2008). Over the last 150 years, observatories around the world have recorded a significant decrease in the global dipole moment and triggered a discussion around whether a reversal is imminent (e.g. Constable and Korte, 2006;Olson, 2002). Such a reversal or extremes of secular variation could have dramatic effects on our modern lifestyle. Knowledge about the field's evolution will help place recent or historic field variations in a long term and geological context and to predict the geomagnetic field's future behaviour. The creation of a detailed record of field variations throughout the last 15,000 years in New Zealand is the major aim of this thesis.
The magnetic field is a vector field. At any point on Earth's surface, it has a specific direction and intensity. The field direction is commonly described by two angles. The magnetic declination is the angle measured between geographic north and the horizontal component of the field direction, in easterly direction, and needs to be accurately known in order to navigate using a magnetic compass. The inclination is the angle of the vector below the horizontal plane. Averaged over long time-scales the geomagnetic field appears to be centred about a geocentric axial dipole (GAD).
Currently the best fitting geocentric diople is tilted by 9.7° with respect to the Earth's rotation axis, with its poles, the "geomagnetic poles" at 80.65° N/S respectively (Thébault et al., 2015). More detailed approximations also describe the presence of a notable non-dipolar field. Because of the latter, secular variation is not coherent over long distances, although at nearby locations it follows similar patterns (for example compare Turner et al, 2015b andLillis, 1994).
Even on short time-scales secular variation is apparent, for instance, since the first observations of magnetic declination in New Zealand were recorded by Abel Tasman in 1642, the direction measured in the northernmost tip of New Zealand has migrated from approximately 8° to 20° easterly declination (Thébault et al., 2015;Van Bemmelen, 1898).
The main source of Earth's magnetic field lies within the fluid, iron-rich outer core of the planet, where convection in combination with Earth's rotation drives a self-sustaining magneto hydrodynamic process -the geodynamo. Numerical simulations of such a dynamo (e.g. Glatzmaier and Roberts, 1995) have reproduced typical features of the Earth's magnetic field i.e. the dipolar nature, secular variation, excursions and polarity reversals. To improve such simulations, we require datasets and models that describe the field evolution in the past. Global palaeomagnetic datasets are compiled in a number of databases such as MagIC (http://earthref.org/ MAGIC/) , GEOMAGIA50 (http://geomagia.gfzpotsdam.de/) (Brown et al., 2015;Donadini et al., 2006;Korhonen et al., 2008) and Pint (http://earth.liv.ac.uk/pint/) (Biggin et al., 2009;Biggin et al., 2010).
Spherical harmonic models that approximate the global field have been created for various timeframes and datasets: a) Present day observational data is used to calculate the International Geomagnetic Reference Field (e.g. Finlay et al., 2010;Thébault et al., 2015), b) historical observations were compiled into the gufm model by Jackson et al. (2000) and c) Holocene palaeomagnetic records were assembled into a series of models such CALS3k for the last 3000 years (CALS3k)  and CALS10k for the last 10,000 years (Korte et al., 2011), the latter being recently modified into model pfm9k by Nilsson et al. (2014). The models of the present day and historical field are constrained by data from modern observatories and satellites or other historical observations. To go further back in time we need to study the magnetic remanence recorded within fired archaeological artefacts or natural materials such as rocks and sediments. Lava flows, for instance, acquire a thermoremanent magnetisation (TRM) upon cooling through the Curie temperature of the constituent ferro-/ferrimagnetic minerals and provide instantaneous records of the palaeomagnetic direction and absolute intensity. The palaeointensity and direction extracted is essential for the calibration of continuous PSV records from sediments, which provide relative estimates of the palaeointensity only, often suffer a depositional inclination error and are subject to amplitude smoothing. Vice-versa, the ages of the lavas can be refined by comparison of their palaeomagnetic information with the continuous records (Lanos, 2004;Pavón-Carrasco et al., 2011;Speranza et al., 2008). Similar processes and methods apply for fired archaeological materials. While archaeomagnetic datasets are restricted to times of human settlement, which in New Zealand reaches back to ~700 yrs BP, dated lavas provide records throughout the volcanic history of an area and hence in the range of thousands of years.
In the southwest Pacific region Earth's magnetic field displays several interesting features. Its proximity to the magnetic South Pole, the point on Earth's surface where the field is vertical, places it into the vicinity of highly dynamic patches of intensive magnetic flux. These are distributed about the northern and southern hemisphere and have not yet been fully explained (Constable et al., 2000;Korte and Holme, 2010). In addition, statistical studies of PSV throughout the last 5 millennia have revealed a very low level of secular variation across the Pacific (e.g. Johnson and Constable, 1998;Walker and Backus, 1996). Records of Holocene PSV are required to re-construct the evolution of both features, however all current global models and datasets suffer from a paucity of data from the southern hemisphere and the South West Pacific in particular. For instance, until recently only one continuous record spanning the last 2500 years was available from New Zealand (Turner and Lillis, 1994) and few more records from Australia (Barton and Mc Elhinny, 1982;Constable and Mc Elhinny, 1985). During recent efforts to better constrain Holocene PSV in the region, archaeomagnetic data have become available on Pacific island pottery (Stark et al., 2010) and New Zealand hangi stones (Kinger, in prep). In addition Turner et al. (2015b) published a new lake sediment record from the South Island in New Zealand, spanning the last 11.5 ka, and efforts to construct a PSV mast r curve for New Zealand are under way (Turner et al., 2015a). Age control has been a limiting factor during previous PSV studies on volcanic materials in New Zealand. Early palaeomagnetic studies, carried out at a time or in an area where no or little age information on volcanic materials was available, approached secular variation from a statistical perspective (Cox, 1969;Tanaka et al., 1997). Other studies were carried out with the purpose to date or provide chronological constraints on the sampled materials by comparison with PSV reconstructions (Downey et al., 1994;Robertson, 1986) or in older lava sequences with periods of reversed polarity (Tanaka et al., 1996). The establishment of eruptive stratigraphies for the main rhyolitic volcanic centres in New Zealand (Froggatt and Lowe, 1990;Nairn, 2002;Wilson, 1993) and radiometric dating of andesites from the Tongario Volcanic Centre (Conway et al., 2016;Gamble et al., 2003;Hobden et al., 1996) allowed the production of the first temporally defined discrete datasets (Tanaka et al., 1994;Tanaka et al., 2009), and enabled comparison with continuous datasets or PSV data from elsewhere (e.g. Turner and Lillis, 1994).
Palaeomagnetic data was also obtained during studies that were carried out with the aim to constraint emplacement temperatures of volcaniclastic materials (McClelland and Erwin, 2003;McClelland et al., 2004). However the distribution of palaeomagnetic data is still sparse and many of the associated age controls have since been revised (e.g. Lowe et al., 2013). Consequently the available datasets require critical re-assessment.
Absolute palaeointensity data is urgently required to constrain the new lacustrine records and to reconstruct fluctuations in the global dipole strength throughout time. However to date only four absolute palaeointensity estimates on volcanic products were published for the last 15,000 years of secular variation in New Zealand (Tanaka et al., 1994;1997;2009). Absolute palaeointensity studies require ideal sample material, i.e remanence carriers need to be single domain grains and thermally stable, a requirement that is difficult to satisfy (Biggin et al., 2007), and this is probably the source for low success rates found during previous studies (e.g. Tanaka et al., 2009).
New radiometric and other age controls have become available during an ongoing, multidisciplinary research programme on the geological and geochemical evolution of the Tongariro Volcanic Centre (TgVC) in the central North Island, New Zealand (e.g. Conway et al., 2016;Eaves, 2015;Gamble et al., 2003;Tost and Cronin, 2015;Townsend et al., in prep). For the first time, the glacio-volcanic history of Ruapehu volcano was approached as well, with a focus on the last glaciation preceding the Holocene (Eaves, 2015). Radiometric dating of young lava flows is difficult due to the long half-life of 40 Ar, the low potassium content of the rock samples, and accordingly, the eruption ages provided have uncertainties in the range of several thousands of years. However, the location of the lava flows within known glacial moraines, together with stratigraphic relationships of chosen samples, have enhanced confidence in the ages provided (Conway et al., 2016). An additional refinement of some ages can be achieved using palaeomagnetic correlation techniques (e.g. Lanos, 2004).
This research programme, and also a recent revision on the tephra stratigraphy of eruptive episodes in the wider North Island (e.g. Lowe et al., 2013) have provided the necessary framework for the palaeomagnetic study on Holocene volcanic rocks presented in this thesis.

Thesis goals
This thesis has three major goals: 1) To built a sequence of discrete palaeomagnetic secular variation records from young volcanic rocks from the Taupo Volcanic Zone in New Zealand, with a focus on the Tongariro Volcanic Centre.
2) To review previously published discrete PSV data from the region and integrate all datasets with continuous records of palaeomagnetic direction and relative intensity from lake sediments (Turner and Lillis, 1994;Turner et al., 2015b) and global field models (e.g. Nilsson et al., 2014). This will integrate with the on-going development of a PSV Master Curve from New Zealand (Turner et al., 2015a).
3) To refine some of the independent radiometric or stratigraphic ages on lava flows by comparison of our new palaeomagnetic data with existing, well dated records.

Thesis structure
This thesis is presented in seven chapters, including three chapters that are specifically written for publication in international journals.
Chapter 3 is based on a manuscript that has been accepted for publication in the "Geophysical Journal International". Chapters 4 and 5 cover specific themes of the research programme and will be edited for publication in due course. In all cases, the contributions of individual authors, the respective journals and a list of relevant appendices are stated at the beginning of each chapter. Additional chapters that introduce context and the applied methodology are provided. Some overlap, in particular within the introductory, geological settings and methods sections was unavoidable. Such overlap is in particular present between chapters 3 and 4.
Both chapters are based on results from a sampling campaign from the TgVC, and geological setting, rock magnetic findings and site descriptions detailed in chapter 3 are represented in tabular form in chapter 4.
Chapter 1 (this chapter) provides an introduction and overview of the main aims of this thesis.
Chapter 2 outlines the palaeomagnetic and rock magnetic methods used in this study.
It introduces the sampling procedures and sample orientation in the field and the subsequent preparation in the laboratory. The methods used for palaeomagnetic demagnetisation and intensity experiments, as well as rock magnetic analysis is explained.
Chapter 3 is a study on the palaeomagnetic directions recorded by lavas from the Tongariro Volcanic Centre (TgVC). The geological setting and the sampled flows and sites are introduced. An overview of the rock magnetic behaviour is given to the extent required for the interpretation of the palaeomagnetic (directional) data. Subsequently the behaviour during demagnetisation experiments and the following statistical data treatment are described in detail. The new data is compared to a recently published continuous PSV curve for New Zealand (Turner et al., 2015b). A comparison is also used to refine the age controls on five of the sampled flows and thus provides constraints on the eruptive history of the volcanoes of the TgVC.
Chapter 4 is a comprehensive palaeointensity study on the lavas introduced in chapter 3. The results of both thermal and microwave palaeointensity experiments are presented together with a detailed investigation of the magneto-mineralogical composition and properties of the samples. The final discussion is two-fold: The new palaeointensity data is discussed in the context of global and regional field-models, and the difficulties arising during thermal and microwave palaeointensity experiments are drawn out in the context of rock magnetic properties and the oxidation state of the iron-titanium-oxides for individual sampling sites.
Chapter 5 presents new palaeomagnetic results from a selection of Holocene rhyolitic lava flows and domes from the Taupo Volcanic Zone, with a focus on the Okataina Volcanic Centre. Palaeomagnetic data on some of the lavas sampled was previously published in two studies by Tanaka et al. (1994;2009). The differences and similarities between the original and the new data are discussed.
Chapter 6 presents a compilation and critical evaluation of all palaeomagnetic datasets on Holocene volcanic materials from New Zealand. The revised data is integrated with the new data from this thesis and compared to sedimentary records from Lake Mavora in Fiordland, NZ (Turner et al., 2015b), Lake Pounui in the southern North Island, NZ (Turner and Lillis, 1994), and a reconstruction of the variation in Earth's virtual axial dipole moment (VADM) by Knudsen et al. (2008).
The effect of the new data on the most recent field model pfm9k (Nilsson et al., 2014) is demonstrated.
Chapter 7 summarizes the main findings of this thesis. a) Sun-compass programme and representative in-and output files. File suncompass-README.txt contains user information and an explanation of the program. Path: Suncompass/ suncompass-README.txt and associated files b) Comparison between the new sun-compass program and an online calculator provided by the US National Oceanic and Atmospheric Administration. The azimuth of the sun was calculated for 30 random locations, time and dates between the years 2011 and 2015 (file format: *.xlxs) Path: Suncompass/ Suncompass-comparison.xlsx

Introduction
This chapter gives an overview of the palaeomagnetic and rock magnetic methods used, including the sample collection in the field. The data analysis and selection criteria applied to the palaeomagnetic data are detailed in chapters 3 to 5. A full description of the physical background to palaeomagnetic and rock magnetic techniques is beyond the scope of this thesis. Suitable introductory texts are Tauxe (2015) and Butler (1998). Dunlop and Ӧzdemir (1997) provide a detailed treatment of the rock magnetic theory.
For summaries of different Thellier-type palaeointensity methods and the corresponding data analysis please refer to Tauxe and Yamazaki (2007) and Leonhardt et al. (2004a). Background to the theory, development and application of the microwave method are given in Hill and Shaw (1999) and Suttie et al. (2010) and references therein.

Sampling and sample preparation
The primary aim of this thesis was to extract reliable palaeomagnetic directions and intensities from the natural remanent magnetisation (NRM) recorded in lava flows and domes from the Taupo Volcanic Zone. It was therefore important to collect accurately oriented samples. Each lava flow or dome was sampled in at least three sites, accessibility permitting. At every site 4 -10 cores with 2.5 cm diameter and 6 -12 cm length were drilled using a petrol driven and water cooled rock-drill. To reduce the environmental impact remote locations were chosen and outcrops were usually drilled at a maximum of one hand-length above the ground. All cores were given a specimen internal coordinate system and oriented using magnetic and sun-compass bearings wherever possible prior to removal from the outcrop.
The specimen internal coordinate system applied in this study corresponds to a standard Cartesian system, the z axis pointing along the core-axis, the y axis corresponding to the horizontal in right handed direction and x pointing vertical to z and y along the upper face of the cores (Figure 2-1c). In order to reference the sample coordinate system to geographic coordinates, we determined the plunge of the core axis (z) beneath the horizontal and the magnetic azimuth, which is the angle between y and magnetic north. Sun-compass orientations were determined using a separate orientation device and will be described in detail in section 2.4. After removal the z axis, pointing along the core and tick marks in y direction were marked on each sample ( Figure 2-2). Magnetic compass bearings were corrected for the local declination using the field predictions made by the 2010 International Geomagnetic Reference Field (IGRF) (IGRF calculator: http://www.ngdc.noaa.gov/geomag-web/). Note that after this correction, the orientations calculated from magnetic compass bearings sometimes differed by up to 10° from the sun-compass orientations (section 2.4), which suggests the presence of some small scale anomalies in the magnetic field. Hence when no sun was available we placed the gnomom used for the sun-compass orientation on the orientation device and took a bearing of the sample orientation from a distance between 5 to 10 m. This helped to verify correct orientation of the samples.
At sites that were too remote to drill samples, or if additional sample material was required, we marked the geological strike and dip on the flat surface of block samples and removed them from the outcrop by hand. They were later drilled in the laboratory using a vertical drill press. All samples were cut into 2.5 cm long cylindrical specimens, and the x-axis marked on the outside surface of each specimen. Full sized specimens were used for thermal and alternating field demagnetisation and Thellier-type intensity experiments while core-ends were either crushed for rock magnetic experimentation or prepared for microwave palaeointensity experiments.
Microwave experiments require mini-samples of ~2-3 mm length and 5 mm diameter.
In general I drilled 5-10 long cores from each specimen and cut these into multiple sub-specimens. Individual sub-specimens were trimmed carefully to provide a smooth surface for the vacuum holder system.
After preparation all specimens were stored in zero-field shielding until all palaeomagnetic measurements were completed.
Figure 2-1: Sampling method: a) Author coring an andesitic lava flow on the left and assistant operating pressure sprayer to enable cooling water flow. b) close-up shot drilling. A petrol driven drill with a diamond bit of 2.5 cm diameter is used to core the rock. c) Orienting of a drilled sample prior to removal from the bedrock, the sample coordinate system is indicated in red. Plunge (angle between z and horizontal) and azimuth (angle between magnetic north and y) are measured using an inclinometer, sun-and magnetic compass.  (2014). At some sites we collected oriented hand samples, cores were drilled out of the latter and adjusted to the same coordinate system.

Sample annotation
Throughout this thesis, sample sites are annotated following the system XXyy.n, where 'XX' is the abbreviation chosen for each flow/unit sampled and 'yy' the site numbering. 'n' corresponds to the ID of a sub-site, if applicable. For example site SC01.1 corresponds to subsite 1 from site 01 at Saddle Cone lava flow.
Specimens are labelled in the format XX-zzN-n for cored and XX-zzH-iiN.n for hand samples. 'zz' is the individual core-number counted, starting from 01 for each individual unit. N is an alphabetical character between A and D and deciphers the specimen ID, starting with 'A' from the outermost end of each core/sample. Another number 'n' is added when specimens are further subdivided into multiple samples, for example for microwave or rock magnetic experiments (example: Specimen ID CC12A-1 corresponds to sub-specimen 1 from the outermost specimen from core/sample 12, sampled from Central Crater flow) . Hand samples require an additional numeral, to accommodate the fact that more than one core was drilled from each sample. Here 'zz' corresponds to the number of the hand samples taken from a flow and 'ii' to the i'th core removed from one hand sample.

Sun-compass orientation
Rather than referencing to magnetic north, orientation data obtained using the suncompass is referenced directly to geographic coordinates. The advantage is that local anomalies in the Earth's magnetic field cannot induce an error in the measured orientation. Sun-compass orientation requires accurate determination of the angle between the shadow thrown by a gnomon located on the origin/centre of an orientation disc and the y axis of the specimen coordinate system. In addition the exact position of the sun needs to be known, which is calculated from an accurate record of the time and geographic positon where the measurement was taken.
All relevant calculations were conducted using a Matlab programme, written following the steps outlined in Tauxe (2015), which is based on the recommendations of the 1996 Astronomical Almanac. The programme consists of two Matlab scripts: the function 'suncompass.m' calculates the local declination of the sun (angle between geographic north and the position of the sun) and subsequently the (solar) azimuth or orientation of the sample. This function needs to be included in the path of the Matlab environment. 'Input.m' imports an orientation file including the sampling information, coordinates, local time and the compass declination for all cores sampled per day and creates an output text file including sample orientation calculated using the suncompass in comparison to the orientation obtained by magnetic compass bearings and all other sampling information. Matlab codes, representative in-and output files and all other information required are displayed in the appendices.
To verify the correct operation of the sun-compass it was compared to an online sun calculator provided by the US National Oceanic and Atmospheric Administration (http://www.esrl.noaa.gov/gmd/grad/solcalc/). The azimuthal position of the sun (angle between geographic north and the position of the sun) was calculated for thirty randomly assigned locations on Earth and times between the years 2013 and 2015. The mean difference between the provided Matlab script and NOAA's online calculator was 5*10 -4° and the standard deviation was 0.048° (see appendix). Such difference could be attributed to rounding errors and is insignificant for the purpose of this study.

Rock magnetic investigation
Beside the thermal history of a rock, the rock magnetic properties and mineralogy influence the stability of the natural remanent magnetisation (NRM) carried by a rock sample and are hence invaluable in the interpretation of palaeomagnetic data and design of experimental procedures. The peak field applied during in-field measurements (hysteresis, IRM, backfield and thermomagnetic curves) was 800 mT. All VFTB data were analysed using Leonhardt's (2006) software RockMag Analyser 1.0 and susceptibility data from the Kappa Bridge using Curveval (Chadima and Hrouda, 2012 (MPMS) instruments. The data from these experiments provided a first impression of the rock magnetic properties of the samples, however for consistency they were not included in this thesis and are therefore not discussed further.

Isothermal remanent magnetisation (IRM) and backfield experiments
Ferro-/ferrimagnetic material can acquire a remanent magnetisation without heating when exposed to a field higher than the room temperature coercivity (Bc) of the constituent grains. While the ambient Earth magnetic field is approximately three orders of magnitude smaller than the coercivity of, for example magnetite, a lightning strike can induce a field strong enough to re-magnetise rocks in the field. In the laboratory we can use IRM acquisition as a tool to rapidly investigate magnetic coercivity, remanence and saturation magnetisation by exposing a sample to a strong external magnetic field that is increased incrementally, and measuring the magnetic remanence (Mr) after each treatment step. Upon saturation the IRM acquisition curve flattens out, and the remanence saturation magnetisation (Mrs) can be extracted ( Figure   2-3). Mrs and the applied field required to saturate a sample are related to the magnetic mineralogy.
In most experimental procedures IRM measurements are followed by backfield measurements, during which the experimental procedure is repeated in a field of opposite direction. Starting at saturated state, an antiparallel field of strength Bcr is required to reduce the remanence back to zero (Figure 2-3). Bcr is also called the coercivity of remanence.

Magnetic hysteresis
During hysteresis experiments a sample is exposed to a strong magnetic field that is increased from zero to a strong field Bmax, followed by a decrease to zero and reapplication of a field in the reverse direction. Subsequently the field is decreased to zero and increased to Bmax again. In contrast to IRM experiments, during hysteresis experiments the magnetic moment is measured continuously during application of the field and it thus measures ferri-/ferromagnetic remanence carriers as well as diamagnetic, paramagnetic and/or superparamagnetic particles. Ferri-/ferromagnetic behaviour produces a hysteresis loop, characterised by its opening, which is defined by the magnetic coercivity Bc, and the saturation magnetisation Ms. Bc can be read of the intersection of the descending and ascending hysteresis curves with the B-axis ( Figure 2-4). Paramagnetic behaviour produces a linear increase of the magnetisation with applied field. Diamagnetism is negative and in the presence of ferri-/ferromagnetic and paramagnetic particles is negligible, superparamagnetism produces a similar effect to the ferri/ferromagnetic particles but with an infinitely small coercivity.
Volcanic rocks usually contain an assemblage of grains that behave differently to each other. The dashed plot shown in Figure 2- Tauxe et al. (1996).

Magnetic domain state
The magnetic domain state is primarily defined by grainsize. Knowledge of the latter is essential for sample selection and interpretation of palaeomagnetic data: In contrast to single domain (SD) grains, multi domain (MD) grains react to an applied field by domain wall movement rather than magnetic moment flipping (Levi, 1977;Néel, 1955). Domain wall nucleation requires a lower energy, and MD grains thus exhibit relatively low coercvity and unstable remanence. The width and steepness of hysteresis loops and backfield curves and hence the ratios Mrs/Ms and Bcr/Bc are related to the magnetic domain state (Dunlop and Ӧzdemir, 1997). Throughout this thesis I display plots of the two ratios for each hysteresis loop against each other, a diagram first suggested by Day et al. (1977). Day et al. (1977) Dunlop (2002): SD magnetite or titanomagnetites exhibit the ratios Mrs/Ms > 0.5 and Bcr/Bc < 2, to qualify as predominantly MD a lower ratio Mrs/Ms < 0.02 and a higher ratio Bcr/Bc > 5 are required. Dunlop (2002) also showed that mixtures between SD and MD grains, often found in natural samples, plot within the PSD section of the diagram along near hyperbolic curves.
Throughout this thesis I display Day-diagrams including Dunlop's (2002) boundaries and representative mixing curves. The reader is advised that while these boundaries and mixing curves provide a reference frame, the exact grain size distribution cannot solely be defined from this diagram.
If adequate instrumentation is available, the measurement of First Order Reversal Curves (FORC) may reveal further information about individual coercivity populations and magnetic interaction within a sample (e.g. Roberts, 2000). These experiments involve repeated measurement of hysteresis loops from a variable starting field but have not been carried out in this study. In addition to single heating and cooling curves, a selection of samples was also subjected to repeated heating χ vs. T experiments with the aim of identifying the onset temperature of thermal alteration. These experiments included heating samples to stepwise higher temperatures and subsequent cooling during the measurement of susceptibility curves.

Thermomagnetic experiments
Curie points were estimated from Ms vs. T graphs using the second derivative approach (Tauxe, 1998) and picked from χ vs. T curves at the onset temperature of linearity in 1/χ , which, following the Curie-Weiss law corresponds to paramagnetic behaviour (Petrovskӯ and Kapicka, 2006) (Figure 2-5). It is noteworthy however that neither method works well in estimating the lower Curie temperature(s) in samples carrying more than one ferrimagnetic phase. In Ms vs. T curves such lower Curie temperatures are often not expressed well, while application of the 1/χ method to χ vs. T data presupposes that above the Tc behaviour is purely paramagnetic, which is not true if there is another higher Curie temperature. In addition, thermally induced magneto-mineral alteration can cause sharp drops in Ms and/or χ that may be misinterpreted as Curie temperature.

Imaging methods and quantitative analysis
The rock magnetic experimentation discussed above reveals a variety of information about the type, magnetic behaviour and thermal stability of the carriers of magnetic remanence and other magnetic minerals in a sample. However, it is difficult to identify secondary mineral phases, formed for example by low temperature or deuteric oxidation solely based on the rock magnetic results. Imaging methods allow rapid petrographic description and identification of different mineral phases. However the threshold grainsize needed to record a magnetic remanence is usually below the resolution of optical methods. For instance a SD magnetite has a maximum size of 0.3 μm (Dunlop and Ӧzdemir, 1997) and best results can be achieved by combining both techniques.
In this study I conducted detailed petrographic imaging on samples from the

Petrographic microscopy
Polarisation microscopy uses the variation in the optical properties of different crystal structures under plain polarised and cross polarised transmitted light for mineral identification and allows instantaneous petrographic description of the main rockforming minerals and texture. Spinel and orthorhombic phases, found in this study are opaque and were therefore analysed under reflected rather than transmitted light.
Mineral identification under reflected light is primarily based on observations such as shape and the reflectivity of the minerals examined. The main magnetic mineral phases identified during petrographic analysis in this study were intergrown phases of the cubic-centred titanomagnetites and rhombohedral hemo-ilmenites (refer to chapter 4 for details). Titanomagnetites are isotropic, and in contrast the rhombohedral phases show low bireflectance. The rhombohedral phases could therefore be distinguished from the cubic minerals under plane polarized light -the amount of light absorbed within the rhombohedral crystal structure and hence the shading of the grain changes with the orientation of its optical axes in the path of the light, which can be adjusted by rotation of the objective table.

Electron microprobing
Rather than relying on the transmitted or reflecting properties of light, during microprobing samples are subjected to a focussed scanning electron beam and detailed imaging is possible from the emission of backscattered electrons (BSE) and secondary electrons (SE), while quantitative geochemical analysis can be done using energy dispersive x-ray spectroscopy (EDX) or wavelength dispersive spectroscopy (WDX).
In this study I used BSE imaging in combination with EDX for first identification of oxide mineral phases. BSE imaging is based on the backscattering properties of the probed grains, the number of backscattered electrons is linked to the atomic mass of the elements the electron beam collides with. Iron and titanium are of higher atomic mass than the constituents of most rock forming minerals and stand out as bright spots in a BSE image. By adjusting the brightness and contrast of the image sufficienty, resolution can be achieved for imaging of exsolution textures between (titano)magnetites and (hemo)ilmenites in grains sized > 5μm. In such images, the (hemo)-ilmenites appear darker due to the lower atomic mass of titanium than that of iron. EDX analysis measures the characteristic spectra from x-rays diffracted from different elements at a spot probed and helps to rapidly identify mineral phases. The composition of a selected number of iron-titanium oxides was quantitatively determined using WDX analysis. The latter method is based on the same principle as EDX, however by isolating between the characteristic wavelengths it detects one predefined element at a time and allows more precise calibration and determination of compositions. But it requires prior selection of synthetic crystals that have specific lattice parameters for the diffraction of the wavelengths of interest. The electron microprobe at Victoria University is equipped with five wavelength dispersive x-ray spectrometers, which permits quantitative analysis of up to 10 preselected elements within the same experimental run. Beam current and diameter can be adjusted to the target properties. Careful calibration with synthetic crystals or natural materials of known composition is a prerequisite. Details of the calibration procedure and beam properties applied in this study are given in chapter 4.  (Butler, 1998). In thermal demagnetisation samples are heated and cooled in zero-field and subsequently measured. A more rapid approach used is alternating field (AF) demagnetisation. Here samples are exposed to an alternating field, the amplitude of which is smoothly decreased from a preselected peak value to zero. This results in a randomisation of the magnetic moments with coercivities lower than the peak field.

Palaeomagnetic direction
AF treatment is applied to all three orthogonal axes of a sample.  (Chadima and Hrouda, 2006;Kirschvink, 1980). The software also allows the user to switch between the specimen coordinate system (X, Y, Z) and the geographic reference frame (N/S, E/W. up/down). Site and flow average directions and associated 95% confidence cones (α95) were calculated following the procedures outlined for statistics of spherical distributions of directions by Fisher (1953).

Introduction
Determining the absolute strength or intensity of the ancient geomagnetic field is one of the major challenges in palaeomagnetic research. A number of different approaches have been proposed in the past, and the most frequently used are based on the Thellier method (Thellier and Thellier, 1959). Throughout this thesis, palaeointensity estimates were made using a combination of thermal and microwave based methods. Thermal experiments were carried out at the University of Liverpool

Concept and palaeointensity protocols
In all methods, the palaeointensity Banc is estimated by assuming a linear relation between the thermoremanent magnetisation (TRM) and the field in which it was acquired. By comparing the NRM with a TRM grown in a known field in discrete intervals of the blocking temperature spectrum, the palaeomagnetic fieldstrength can be found using: Where │ΔNRM│ is the magnitude of NRM lost between two demagnetisation steps and │pTRM│ is the magnitude of the partial TRM (pTRM) gained in a laboratory field between the two temperatures. In practice the method consists of three steps: Rock samples are stepwise heated from room temperature to temperatures close to the Curie point of common minerals, each temperature interval includes one heating and cooling cycle in a zero-field and one in a known stable field. The remanences lost and the pTRMs obtained in each cycle are calculated and displayed in respect to each other in a so-called Arai diagram (Figure 2-7). Banc can be calculated from the best fitting slope to the data. For relation (1) to be valid the behaviour of the respective sample needs to obey three fundamental principles, based on the theory of single domain grains (Néel, 1955) : (1) Independence: The pTRMs acquired after each heating step are independent of each other.
(2) Additivity of pTRM: The total TRM gained is equal to the sum of all pTRMs gained during individual temperature steps.
(3) Reversibility: The unblocking temperature Tub equals the blocking temperature Tb for any magnetic carrier in the sample.
To obey this, the NRM must be a TRM. Further the magnetic carriers within a sample need to be thermally stable, non-interacting and of single-domain size. These requirements are rarely satisfied in nature. Titanomagnetites and titanohematites, the most prominent magnetic mineral phases found in samples of this study (chapter 3-5), are highly reactive when subjected to heating and the onset of thermal alteration is often observed within the temperature range applied during most standard Thellier experiments. Furthermore it has been shown that the reversibility and independence criteria are not satisfied in the presence of MD grains (Fabian, 2001;Levi, 1977;Shcherbakov et al., 2001). To detect non-ideal behaviour, a number of checks and modifications to the basic method have been proposed (e.g. Coe, 1967;Riisager and Riisager, 2001;Yu et al., 2004). In this thesis I applied the two most frequently used modifications, the Coe and IZZI type protocols: The conventional Coe type protocol (Coe, 1967) involves double heating steps, each temperature interval starting with the zero-field step, followed by an infield step. After every second in-field step an additional zero-field or tail check (Riisager and Riisager, 2001) was implemented. This check shows whether remanences acquired during the preceding pTRM check are adequately removed in the following zero-field steps and hence monitors the reversibility between Tb and Tub. Each tail check was followed by a pTRM check during which an earlier in-field step is repeated. This check helps to identify changes to the ability of a sample to retain a pTRM caused by thermo-chemical alteration. The IZZI protocol (Tauxe and Staudigel, 2004) involves double heating steps, alternating in order of zero field (Z) and infield (I) steps. pTRM checks are included between the zero-field steps. The protocol does not require tail checks: pTRM tails carried by MD grains would result in a systematic difference between the zero-field remanences measured during zero-field first (ZI) and in-field first (IZ) heating intervals.
In order to ensure that only reliable palaeointensity results are used for the calculation of a flow mean, in this thesis the data were subjected to stringent selection following a set of criteria that assesses the quality of the slope fit within the Arai plot, pTRM and tail checks. Different sets of selection criteria have been proposed in the past and often these were somewhat adjusted to the overall quality of the data. It is important to note that while too 'soft' criteria sets can result in acceptance of erroneous palaeointensity results, overly stringent sets can result in removal of useful data, where scatter is due to instrumental noise (e.g. Paterson et al., 2014).
For a more thorough discussion of the palaeointensity methods employed and the selection criteria used please refer to chapter 4. . The data shown includes passed and failed data points, pTRM and tailchecks. The │TRM│ gained and │NRM│ remaining are normalized against NRM0. Thermal alteration at temperature steps higher than 400°C results in a change in slope. These points, displayed in grey have been rejected from further analysis. The palaeointensity can be calculated from the best fitting slope.

Microwave palaeointensity
The microwave technique (Hill and Shaw, 2000) was designed to minimize alteration by exciting the magnetic moments of the ferro-/ferrimagnetic grains using energetic microwaves instead of the heating-procedure, while otherwise using methods similar to those used for thermal palaeointensity determination. Heating of a sample results in lattice vibrations that excite spin waves and create magnetic moment. In contrast, high frequency microwaves can directly excite the electron spins without excessive heating of a sample, a technique known as ferromagnetic resonance (FMR) (Hill and Shaw, 1999;Walton et al., 1996). Microwave exposure times are usually between 3 and 15 seconds and the method is therefore significantly faster than thermal applications. This allows single specimes to be treated individually and enables better selection of adequate treatment steps than achieved during bulk heating experiments. To date, no theoretical relation has been established between microwave and thermal energy, but a number of comparative studies prove the method reliable (e.g. Hill et al., 2002).
Microwave palaeointensity experiments were conducted using the system at the University of Liverpool. The central part of the system is a microwave demagnetisation/re-magnetisation unit that is coupled to a Tristan Technologies, Inc.
DRM300 cryogenic magnetometer (Suttie et al., 2010). The magnetometer has a maximum sensitivity of 10 -11 Am 2 and needs to be fine-tuned prior to each experiment.
The microwave unit consists of a signal generator which produces a signal of frequency between 14 and 14.5 GhZ and is connected to a resonance cavity along a horizontal brass tube, which acts as waveguide. The cavity is shielded from external magnetic fields and contains three field coils creating magnetic fields of pre-defined strength and direction during in-field steps. Specimens of ~ 5 mm diameter and up to 3 mm length are mounted on a vertical quartz glass tube using a negative pressure system that allows automated movement of the sample between the microwave cavity, for demagnetisation or growth of a microwave induced remanence (TMRM), and the magnetometer for subsequent measurement. During microwave application the sample is placed in the centre of the cavity, at the node of minimum electric but maximum magnetic energy (Suttie et al., 2010). The resonance properties in the microwave cavity are dependent on the sample size and its position in the cavity. Prior to an experiment the exact frequency (range 14-14.5 GhZ) is fine-tuned by frequency sweeping in order to achieve maximum energy absorption rates.
The total energy applied to a sample is the integrated amplitude of the microwaves and exposure time, demagnetisation steps may be selected by increasing either the exposure time or the power of the microwave source. After each step the absorbed energy is plotted against the reflected energy. In an ideal case the total energy applied would be absorbed by the sample. Small changes to the positioning of the sample in the cavity or alteration to the magnetic minerals can result in changes. This can result in failure of Coe and IZZI type experiments, which require the total power absorbed between repeated treatment steps to remain consistent. Further it is noteworthy that although in theory no heating of the specimens is to be expected, di-electric heating has been found to result in heating of the bulk sample up to 250°C (Suttie et al., 2010).
During the experiments carried out for this thesis I also identified small melt spots on a few samples. Similar observations were made for example by Stark (2011), suggesting that spot heating can be of much higher temperature.

Abstract
The paucity of Holocene palaeomagnetic data from the southern hemisphere significantly limits the resolution of global models of palaeosecular variation. To date, studies on lava flows from the southern hemisphere, and New Zealand in particular, have been significantly hampered by a lack of adequate radiometric age control. We sampled 39 new sites on twelve Holocene basaltic andesite and andesitic lava flows from the Tongariro Volcanic Centre in the central North Island, New Zealand. Formation ages for these flows are constrained by new high-precision

Introduction
Palaeomagnetic studies of volcanic rocks contribute both to our understanding of geomagnetic secular variation and to the eruptive history of the volcanic area studied.
The palaeomagnetic information contained in the thermoremanent magnetisation (TRM) of lava acquired during cooling provides an instantaneous record of the palaeofield. Comparison of the palaeomagnetic direction and intensity recorded in a lava flow with a palaeosecular variation master record or model for the studied region can help to constrain its eruption age more precisely.
Understanding palaeosecular variation (PSV) on time-scales of hundreds to thousands of years is an important step in understanding the origin of Earth's magnetic field and the prediction of its future behaviour. However, current global field models (e.g. Korte et al., 2011;Nilsson et al., 2014) suffer from a paucity of data from the southern hemisphere. For instance, until recently only one continuous record, spanning the last 2500 years, was available from New Zealand (Turner and Lillis, 1994), the nearest other records being from Australia (Barton and Mc Elhinny, 1982;Constable and Mc Elhinny, 1985). Several attempts have been made to study PSV recorded in New Zealand volcanic rocks (Cox, 1969;Downey et al., 1994;Robertson, 1986;2007;Tanaka et al., 1994;1996;1997;2009) but these studies were constrained by the limited geological and/or age information available at the time. During a recent geologic mapping programme of the Tongariro Volcanic Centre (TgVC) (Conway et al., 2016;Townsend et al., in prep), new 40 (Conway et al., 2016;Jicha et al., 2012). However, such results are still subject to uncertainty in the range of hundreds to thousands of years, which is significant for Holocene-aged lavas.
Here we present the palaeomagnetic results on 12 Holocene lava flows from Ruapehu and Tongariro volcanoes. Eruption ages for the sampled flows are constrained by highprecision 40 Ar/ 39 Ar geochronology (Conway et al., 2016), tephrochronological or geomorphological information. The primary aim of this study is to develop a series of discrete secular variation records for this area for the past 15 kyrs. Comparison of our new data with a recently published lake sediment-based PSV record for New Zealand (Turner et al., 2015b) allows refinement of the existing age control of the lava flows.
Our new data will significantly improve global and regional field models. In the long term we aim to establish a comprehensive dataset and field model for the southwest Pacific region, which will contribute to an understanding of the origins of the geomagnetic field and will also provide a powerful new dating tool for young volcanic deposits across the region.

Geological Setting
The Tongariro Volcanic Centre (TgVC) is situated at the southern end of the Taupo Volcanic Zone, an area of geothermal and volcanic activity that extends as far north as White Island in the Bay of Plenty (Wilson et al., 1995) (Figure 3-1, inset). The TgVC comprises four main volcanic edifices: Kakaramea, Pihanga, Tongariro, and Ruapehu (Cole, 1978;Hobden, 1997). It is dominated by Mt Ruapehu at the southern end and a complex of smaller andesitic cones around Mt Tongariro (Hobden et al., 1996) on the northern end. The two main cones, Mts Ruapehu and Tongariro, are large and complex intermediate arc composite volcanoes with eruptive histories stretching back to at least 340 ka (Hobden et al., 1996;Tost and Cronin, 2015).
Pioneering work on the volcano stratigraphy of Mt Ruapehu was carried out by Hackett (1985), who defined four major formations (from oldest to youngest: Te Herenga, Wahianoa, Mangawhero and Whakapapa), which represent the deposits of episodes of effusive eruptive activity that w interrupted by periods of erosion and relative quiescence. Deglaciation of the edifice since ca. 15 ka affected the distribution of lava flows due to the retreat of former large flank glaciers that had occupied valleys for at least the preceding ca. 35 kyr (Conway et al., 2015). Thus, post-glacial lava flows, which are assigned to the Whakapapa Formation (<15 ka), can be identified primarily from their location within glaciated valleys and also the lack of evidence for large-scale lava-ice interaction or significant overlying moraines (Conway et al., 2015). Until now, absolute age control for flows belonging to the Whakapapa Formation has been sparse and prone to large uncertainty (Gamble et al., 2003). In addition to 40 Ar/ 39 Ar eruption ages presented by Conway et al. (2016) and used here, chronological constraints are provided by tephra and volcaniclastic marker beds from the well-dated suite of airfall deposits that have been catalogued in high-resolution studies of the ring plain stratigraphy (Donoghue et al., 1995;Donoghue and Neall, 2001;Moebis et al., 2011). Volcanic activity since 2 ka at Mt Ruapehu has primarily been of a phreatomagmatic nature (Price et al., 2012). Detailed descriptions of the stratigraphy and morphology around Mt Tongariro shows a wide array of young lava flows (Hobden, 1997). Although no direct radiometric dates for these young flows are available, their minimum ages are constrained by overlying tephra marker beds with associated radiocarbon dates (Hobden, 1997).

Samples and ages
Field sampling for palaeomagnetic analysis was carried out over two seasons from 2012 to 2014, alongside a geological mapping programme of the Tongariro National Park undertaken by GNS Science (Townsend et al., in prep). Holocene lava flows were selected on the basis of new 40 Ar/ 39 Ar eruption ages (Conway et al., 2016), and dating using available tephrochronology (Hobden, 1997;Topping, 1974) and morphological features (Conway et al., 2015).
Altogether we sampled 39 sites from 12 individual flows (Figure 3-1). The morphology and exposure of the individual flows vary considerably and are discussed below. Each flow was sampled at a minimum of three sites along the flow, accessibility permitting. The flow tops are usually strongly auto-brecciated into smaller blocks and they were therefore sampled at multiple sub-sites to verify that individual blocks were in situ. At every site or sub-site 4-10 cores of 2.5 cm diameter and 6-12 cm length were drilled using a water-cooled, petrol driven hand drill. All cores were oriented using a magnetic compass, and sun-compass wherever possible. Magnetic compass bearings sometimes differed by several degrees from the sun-compass corrected orientations and so, when no sun was available, we took a bearing from a distance of ca. 10 m to verify correct orientation of the samples. At sites that proved unsuitable for drilling, hand samples were obtained following the marking and measuring of a suitable flat surface for later re-orientation. These were then drilled in the laboratory using a vertical drill press. All samples were cut into standard sized specimens. The sampling details for all twelve flow units are presented below. All Ruapehu flows studied belong to the Whakapapa Formation (Hackett, 1985;Townsend et al., in prep), and are listed together with the age information in   Note: Abb is the abbreviation for the flow/ unit name. The member corresponds to the classification revised in Townsend et al. (in prep). Central Crater flow (CC) is associated to Mt Tongariro. Age is the age control available for the respective flows from different sources with 2σ standard error, if the age was obtained by radiometric dating. Site coordinates are listed in Table 2.*Sample ID's for the 40

Whangaehu flow (WE)
This flow is located within a formerly glaciated valley and, with an 40 Ar/ 39 Ar eruption age of 0.2 ± 2.2 ka, it is the youngest dated flow belonging to the Whakapapa  (Conway et al., 2015;Lescinsky and Fink, 2000) and this may also affect the rock magnetic properties of the lava (Biggin et al., 2007;De Groot et al., 2014).
Whangaehu flow was sampled at five sites through its vertical thickness. Sites WE01 (zone A) and WE02 (zone B) are located on a small river terrace that accesses the flow base. Drilled samples were taken from zone A (WE01), while an un-oriented hand sample was taken from a large platey boulder that is assumed to originate from zone B (WE02); the collection of drill cores from zone B was not successful due to the fine-scale platey jointing. Sites WE04 and WE05 are located within zone C. Two oriented hand samples were taken from individual blocks along the flow top (WE03) (zone D). This flow provides a case study for assessing how different cooling rates may affect the rock magnetic characteristics and the palaeomagnetic data.

Bruce Road flow (RP)
The Bruce Road flows are part of an extensive series of a'a lava flows on the lower north-western exposure of the Iwikau member at the Whakapapa ski-field. Topping (1974) suggested an eruption age for one flow between 5 and 10 ka based on the underlying Murimotu Formation and overlying tephra. The flows were sampled at two sites. RP01 is located inside a sharp bend in Bruce Road and RP02 is about 300 m west of RP01, in a steep gorge above Waipuna Stream. At both sites a section through the flow interior is accessible. This is more prominent at RP02 where the Waipuna Stream gorge exposes a vertical section through the flow and the zonation described for Whangaehu lava flow is again apparent. At site RP01 a loose, un-oriented hand sample was taken from the upper surface of the flow, and oriented samples were drilled at two additional sub-sites from the flow interior. RP02 was sampled at two sub-sites within zone C.

Taranaki Falls flow (TF)
Taranaki Falls flow is also part of the Iwikau member and forms a lava flow that is exposed along Wairere stream over a distance of about 3 km. Timing of flow emplacement is constrained by an 40 Ar/ 39 Ar eruption age of 8.8 ± 2.8 ka. We sampled the flow TF at three sites from the rootless, exposed upper end (TF01) to the lower end (TF03) near the "Round the Mountain" walking track, at the top of Taranaki Falls.
Palaeomagnetic sampling of this flow provides a good example of sites spread over a long lateral distance and therefore helps to assess whether sampling location and changes in the flow morphology can affect the recorded directions.

Saddle Cone flows (SC)
Saddle Cone lava flows originate from a small parasitic cone (Topping, 1974) and cover an area of approximately 4 km 2 north of Mt Ruapehu. Topping (1974) proposed a minimum eruption age of 5 ka based on a capping tephra. The flows were later associated with short-lived series of pyroclastic eruptions at around 10 ka based on geochemical and petrographic correlations (Nakagawa et al., 1998). Outcrops include individual blocks along the auto-brecciated flow tops and at the flow fronts. The brecciated nature of the flow surface makes it difficult to distinguish between individual flows. Oriented samples were taken from six sites; at each site we usually sampled several blocks. Altogether 13 individual blocks (sub-sites) were sampled from six sites, dispersed across the area to assess whether significant differences in the palaeomagnetic direction might indicate significant age difference or post-cooling block movement.

Mangaturuturu flow (MT)
The Mangaturuturu flow (MT) sampled in this study is the middle of a sequence of three flows and exposed on a steep bluff within Mangaturuturu Valley. Samples were collected from a single site, from which an eruption age of 11.9 ± 2.2 ka was obtained.
The outcrop comprises mainly fractured, platey lava. We therefore drilled coresamples where possible and otherwise collected hand samples from a wide area across the outcrop.

Skyline Ridge flow (GR)
Two sites were sampled from the eastern side of this narrow, ridge-top flow on the south-western flank of Mt Ruapehu. Due to its location at high elevation and the frequent snow cover it is particularly prone to weathering. Timing of flow emplacement is constrained by an 40 Ar/ 39 Ar eruption age of 15.1 ± 2.4 ka. The flow at the dated location is fractured by platey joints; the remaining exposed parts of the flow are blocky. Site GR01 was sampled at two sub-sites: oriented hand samples were taken from the platey section (zone B, GR01.1) and drilled samples from close by (zone C, GR01.2). Further oriented cores were taken from higher up the flow (zone C, site GR02).

Rangataua flows (RT)
Rangataua andesites are an extensive suite of lava flows on south Mt Ruapehu.
Several stream cuts expose parts of the flow interior; however, the ubiquitous glassy nature of the groundmass precluded 40 Ar/ 39 Ar dating and age controls on these flows are based on field observations only - Price et al. (2012) suggested ages < 9.7 ka for vent-proximal and medial deposits, and > 11.9 ka for distal flows, while Conway et al. (2016) suggests the flows may have been emplaced during the late stages of the last glaciation and could be between 10 and 15 ka old. Altogether we sampled five sites along a distance of 2 km. All sites are located in proximal (RT04-RT05) or medial flows (RT01-RT03), located above the bush-line. Site RT01 is located on a small stream cut immediately beneath the flow top surface at the eastern margin, sites RT02 and RT03 are located within river cuts further downslope. At all three sites drilled samples were collected. Oriented hand samples were taken from the flow top surface at sites RT04 and RT05, but palaeomagnetic direction estimates were only made from the drilled samples. Thermomagnetic experiments included in-field heating of samples to a peak temperature of 700°C with a heating rate of 30°C/min and subsequent cooling. All VFTB experiments were carried out with a peak field of 800 mT. Full analysis of the VFTB data was made using RockMag Analyser 1.0 (Leonhardt, 2006) and of susceptibility data using Cureval 8.0.2 (Chadima and Hrouda, 2012). Curie points were estimated from saturation magnetisation vs. temperature (Ms vs.T) graphs using the second derivative approach (Tauxe, 1998) and picked from susceptibility vs.

Rock Magnetic Properties
temperature (χ vs. T) curves at the onset temperature of linearity in 1/χ, which corresponds to the change from ferrimagnetic to paramagnetic behaviour (Petrovskӯ and Kapicka, 2006).

Results: Hysteresis and IRM
The  (Tauxe et al., 1996). We note that IRM acquisition curves do not saturate at applied fields lower than 300 to 400 mT, which also indicates the presence of a second, high coercivity phase in these

Results: Thermomagnetic experiments
Thermomagnetic experiments on samples from different sites and flows yield a wide range of curve shapes with variable degrees of reversibility. We cannot assign a unique behaviour to individual flows, but we observe a strong dependence on the location within the lava flows sampled. The Curie temperatures range from 150 to 600°C, leading in combination with hysteresis parameters, to the conclusion that titanomagnetites are the primary carrier of magnetic remanence and we relate the differences in the thermomagnetic properties to differing cooling rates and associated oxidation and exsolution states of the Fe-Ti oxides.
We classify the thermomagnetic curves into five general types ( (2002); b) representative hysteresis diagrams. The specimen ID, site ID and hysteresis parameters are displayed in each graph and the corresponding data points in the Day plot labelled from i -vi. Units for Ms, Mrs are *10 3 Am 2 /kg, Bc, Bcr in mT. The hysteresis parameters were calculated using RockMag Analyser 1.0 (Leonhardt, 2006); c) The five general types of thermomagnetic curves identified in this study. See text for details.

Palaeomagnetic methods
Stepwise thermal and alternating field demagnetisation was carried out on up to ten specimens per site and subsequent remanence measurements were made using an AGICO JR6A magnetometer at Victoria University of Wellington or an AGICO JR6 magnetometer at the University of Liverpool. Preliminary thermal demagnetisation (THD) experimentation was carried out on two specimens per site, at temperature intervals of 50°C from room temperature to 600°C or until less than 10% of the initial natural remanent magnetisation (NRM) remained. All samples were subsequently divided into groups of similar behaviour for more detailed demagnetisation experiments. Each subsequent experiment contained ten or more steps of increasing temperature. Two specimens per site were also subjected to progressive alternating field demagnetisation (AFD) at intervals of between 5 and 15 mT up to a peak alternating field of 90 mT. All data were displayed in vector component diagrams (Zijderveld, 1967) and the characteristic remanent magnetisation ChRM directions calculated by principal component analysis (PCA) (Kirschvink, 1980) using Remasoft 3.0 (Chadima and Hrouda, 2006). All heating experiments were accompanied by room temperature magnetic susceptibility checks to identify the onset of thermo-chemical alteration. Site and flow mean directions were calculated using Fisher statistics (Fisher, 1953) following the steps and criteria outlined below: 1) Specimen ChRM directions must include a minimum of five data points with a maximum angular deviation (MAD) of 5°. For specimens in which the ChRM data trend away from the origin in the high temperature/AF demagnetisation steps (possibly due to thermal alteration during the demagnetisation experiment), the ChRM was calculated anchored to the origin.
2) When more than one specimen was demagnetised from a core, their ChRM directions were averaged to give a sample mean. Hand sample averaging was conducted in a hierarchical way: a mean was calculated for each core extracted from a hand sample and all core means were subsequently averaged into a sample mean.
3) Site mean directions were calculated by averaging over the samples from each site.
Data from individual samples were excluded from the calculation of a site mean if their ChRM directions are clearly outliers from the cluster described by other samples from the site and/or there is reason to doubt the result.

4)
Finally, flow mean directions were calculated. In most cases, data were included only from sites for which the site mean direction has an α95 of less than 10°.
Exceptions are site means calculated from three or fewer samples. If the site mean directions agree at the 95% level of confidence, or if their α95 cones overlap and site selection indicates that the samples should record the same direction, it was assumed that the dispersion between the sample data is due to random uncertainty and flow means were calculated by averaging all sample mean directions. At CC however, while the site-mean 95% confidence cones overlap, there is evidence of random syn-cooling block rotation on the flow surface, and a flow mean direction has therefore been calculated by averaging the site means (section 3.7.1).

Demagnetisation behaviour and statistics
In most samples, the demagnetisation behaviour mirrors the rock magnetic properties described above and is related to the thickness and the zone within the lava flow that was sampled. However some sites show anomalous behaviour that requires further discussion. Furthermore the method of averaging varied with the type of sampling carried out; detailed flow by flow descriptions are given below. Site mean and accepted directions for all flows are listed in Table 3

Central Crater flow (CC)
Central Crater specimens have NRM intensities ranging from 18 to 57 A/m and are more strongly magnetized than any of the Ruapehu andesites discussed below.
Specimens from most sites yield univectorial decay to the origin (Figure 3-4a,b) and their ChRM directions form tight clusters. While the 95% confidence cones overlap to some extent for specimens from sites CC02, CC03, CC04 and CC05, we discard all data from CC01 and CC06 due to anomalous directions and/or curved demagnetisation paths, which may have been caused by block movement during and/or after cooling.
Despite the overlap between the 95% confidence cones it cannot be said with certainty that CC02-CC05 are indistinguishable: each block may have rotated, albeit very slightly during the final stages of cooling of the flow interior. On the assumption that any such movement is minor and random between sites, we calculate a flow mean from these four site mean directions, giving a best estimate of Dec = 20.8°, Inc = -67.4°,

Whangaehu flow (WE)
Thermal and AF demagnetisation were carried out on drilled samples from sites WE01 (zone A), WE04 and WE05 (both zone C). Successful results from site WE01 show two distinct blocking temperature (Tb) ranges (100-300°C and 500-600°C) during thermal demagnetisation (Figure 3-4c). During AFD the remanence of a specimen from this site decays rapidly with a median destructive field of 10 mT. The low Tb and low coercivity components carry between 40 and 80% of the total remanence. However, this component also carries anomalous directions and the ChRM has therefore been calculated from the high Tb, high coercivity components.
In contrast, specimens from sites WE04 and WE05 retain 70% of their total NRM until they are heated to a temperature higher than 500°C, and yield straight line decay during demagnetisation (

Delta Corner flow (DC)
Thermal demagnetisation of samples from Delta Corner flow (DC) yields an unusual sequence of normal-reversed-normal polarity components. Similar behaviour was also observed in some specimens from the Whakapapaiti B (WPB) and Skyline Ridge (GR) flows. We attribute it to a partial self-reversal process in a narrow interval of blocking temperatures, as has been reported in other studies of andesiticdacitic materials (Paterson et al., 2010).
The NRM direction of all specimens is of normal polarity, however between 200 and 245°C an antiparallel component of reversed polarity is removed, causing an increase in the intensity by up to 38% (Figure 3-4e, f). Demagnetizing to higher temperatures results in rapid loss of intensity and a ChRM of normal polarity. The reversed showing a lack of correspondence between blocking temperature and coercivity.
A thermal event can be excluded as the cause for the antiparallel component, as if this were the case all grains with blocking temperatures below 250°C would carry a reversed remanence direction. The rock and thermomagnetic causes of partial self-reversal require further investigation and will be reported in a later study.
Palaeomagnetic directions were calculated from the high blocking temperature component (Tb > 250°C). Site DC04, sampled from the flow centre, records a significantly shallower inclination than sites DC01 and DC03, which were sampled from the base and the upper part of the flow. All three sites were collected from the same unit and are assumed to be in-situ based on our field observations. We therefore average all data into a flow mean direction of Dec = 8.5°, Inc = -68.3°, α95 = 3.9°, N = 10.

Bruce Road flow (RP)
Successful ChRM directions and a site average were obtained only from site RP02.
The ChRM directions from this site are well grouped and the NRM usually comprises two components which unblock between 200-300° C and 400-500° C (Figure 3-4h).
Both blocking temperature components are parallel to each other. In contrast, the directions at RP01 display inconsistent declinations and anomalously low inclinations.
The NRM intensities of RP01 specimens are up to two orders of magnitude higher than those of specimens taken at RP02. The demagnetisation behaviour is usually unidirectional, but a number of samples, collected ca. 20 m further along the outcrop on the same site show two components with an overlapping coercivity spectrum ( Figure 3-4g). In the field we observed strong magnetic compass deflection at this site, and note that the site is situated beneath an isolated and exposed hill. We therefore suggest that the anomalously strong remanence and unusual remanence directions were induced during one or several lightning strikes that created short-lived, strong magnetic fields. The strength and directions of the magnetic field induced during a lightning strike is a function of the distance from the location of the strike. Several events in different, nearby locations may have therefore affected different coercivity populations, resulting in the presence of multiple components during demagnetisation.
For this reason all data from site RP01 was discarded.
The flow mean direction includes demagnetisation data and the ChRM directions from three successful Thellier palaeointensity experiments from specimens collected at RP02 only. A full analysis of the intensity data will be reported in a later publication. With the aim of identifying individual blocks that are not in-situ, we have examined the data from every sampled block separately and summarize all successful results into site means. Despite the wide sampling we identified only three blocks from sites SC01 and SC02 that give reliable results. Both sites were located on the south western end of the flow complex, however while SC01 was in an incised stream bed, SC02 comprises two independent blocks on the flow surface, ca. 50 m apart. While the mean direction from the two blocks sampled at site SC02 (SC02.1, SC02.2) agree, the palaeomagnetic direction obtained from sub-site SC01.1 has a more easterly declination, but similar inclination. We therefore calculate two means: SC01.1: Dec = 12.2°, Inc = -58.3°, α95 = 5.7°, N = 4 and SC02: Dec = 358.7°, Inc = -54.9°, α95 = 3.0°,N = 7. We lay higher confidence on the mean from SC02, which includes the ChRM directions from two separate blocks.

Mangaturuturu flow (MT)
All samples show little or no viscous overprint (Figure 3-5a,b). However, specimens demagnetise predominantly either at temperatures below 200°C or at temperatures above 300° C, in general agreement with differences interpreted from their rock magnetic properties. Site statistics were calculated separately for specimens taken from hand samples and cores. As expected, the hand sample data shows larger scatter than the drilled samples; however, distributions from both data sets overlie one another and all data have been included in an overall mean of Dec = 8.7°, Inc = -60.4°, α95 = 4.0°, N =10.
Whakapapapaiti A -C flows (WPA, WPB, WPC)   In contrast, the hand samples taken from platey zone B (GR01.1) carry a much weaker remanence (approximately 0.5 A/m), they demagnetise at generally lower temperatures (300-400°C) and carry low blocking temperature components that result in a slight rise in intensity between 200 and 250° C (Figure 3-5h). This is similar to the behaviour observed in samples from Delta Corner (DC) and Whakapapiti B (WPB) lava flows which show partial self-reversed magnetisation.
The hand sample data was excluded from the calculation of the flow mean due to its high degree of scatter. Our best mean for the flow (Dec = 356.5°, Inc = -81.4°, α95 = 3.8°, N = 11) yields an inclination steeper than expected for this period of time.

Rangataua flows (RT)
Thermal and AF demagnetisation experiments were carried out on drilled samples from sites RT01, RT02 and RT03. All sites show similar THD behaviour, with small viscous overprints (Figure 3-5i). Despite being of single component, some of the thermal demagnetisation experiments show a deflection of the remanence from the origin in demagnetisation steps higher than 400°C, which may be indicative of thermal alteration during demagnetisation. The mean direction for site RT01 has a significantly higher inclination than the mean directions from sites RT02 and RT03, suggesting that the flow sampled at RT01 is of different age. We therefore    A flow mean was calculated from sample ChRM directions from sites RT02 and RT03, which yield overlapping 95% confidence cones. The mean direction from RT01 is distinct from RT02 at the 95% confidence level from the site mean calculated for site RT01, suggesting that the lava sampled at RT01 is of different age.
Note: Flow is the name for each sampled unit as discussed in section 2. Independent age is given in ka, with 2 uncertainty if applicable: the sources are listed in Table 1. Lat and Long are the site geographic coordinates (WGS84). Dec, Inc refer to palaeomagnetic directions. N/n/n0 = total number of samples included in the average/total number of specimens included/ total number of specimens demagnetised. Sample orientations were calculated from sun-compass measurements, unless otherwise indicated. Preferred age is obtained from all available age control including correlation of the palaeodirection with the Lake Mavora PSV reference record (Turner et al., 2015). It is given as a range, and in years BP since the age probability distribution is in general not Gaussian, and the age model of the Mavora record is radiocarbon-based. See text and/or electronic appendices for other details.

Accuracy and precision of the palaeomagnetic data
In Figure 3-6 we plot the site mean directions and associated 95% confidence limits (α95) calculated from all data that passed our selection criteria, and our best estimates of each overall flow mean palaeomagnetic direction. The ChRM directions of all specimens are detailed in the electronic appendix. With flow mean α95's generally less than 5°, the data quality is comparable to or better than that obtained in other palaeosecular variation studies on lava flows (e.g. Speranza et al., 2008;Tanaka et al., 1997). Furthermore, for each flow, all site mean directions agree at the 95% level of confidence, unless otherwise discussed.
Random sources of error associated with field orientation of samples, measurement uncertainties and the estimation of ChRM directions, may be expected to lead to a scatter of 5-10° at a site. With 4-10 specimens per site, this means that an α95 of up to 5° can be attributed to such random uncertainties. With current field and laboratory methods, it would be difficult to further reduce these uncertainties. Possible sources of significant differences between site mean directions for a given flow include syn-or postmagnetisation movement, and small scale magnetic field anomalies resulting in differences in the ambient field between sites at the time of magnetisation. We argue that syn-or post-magnetisation movement is likely and can account for the between-site/ block differences in direction in the cases of the Red Crater flow in Central Crater (CC) and the Saddle Cone flow (SC). In both of these flows we were limited to sampling blocks near the flow exterior which might be prone to such movement.
In the case of Delta Corner flow (DC), the inclination measured at site DC04 is shallower than that recorded at sites DC01 and DC03. The sites are from different zones within the flow and are spread over a lateral distance of 200 m. Calculations (M. Ingham pers. comm.) indicate that an underlying lava flow with a normal magnetisation of 10 A/m could produce anomalies of up to 9° in declination and 3° in inclination 1-2 m immediately above the flow edge, but these figures decrease to as little as 1° within a distance of 20 m. Magnetic field influences of this magnitude are not great enough to explain the differences.
A local or regional anomaly due to a more extensive strongly magnetized underlying body might be expected to manifest in systematic differences between sample orientations derived from sun compass and magnetic compass bearings. Although we observed differences of up to 15°, they were not systematic, and were interpreted as being due to the magnetisation of immediately adjacent rock, i.e. that being sampled.

Comparison with previous New Zealand PSV studies
In Figure 3-7 we display our flow mean palaeomagnetic directions with 95% confidence limits and existing age control, superimposed on the recently published lake sediment PSV record from Lake Mavora, South Island, New Zealand, that covers the past 11,000 years (Turner et al., 2015b). The Lake Mavora record has been migrated to the location of Whakapapa Village (175.54° E, 39.2° S) using a virtual geomagnetic dipole (VGP) transformation (Noel and Batt, 1990 where the gradient of the field is low to moderate. This is well within the uncertainties in the data discussed above. we use palaeomagnetic dating to improve the precision of the independent age control. Not shown on Figure 3-7 is the only other continuous record from New Zealand, a lacustrine record from lake Pounui (Turner et al, 1994), as it has recently been argued that the age model of the record is too old by several hundreds of years (e.g. Turner et al,, 2015b). For similar reasons we do not display the results of the earlier palaeomagnetic studies on volcanic materials from New Zealand or pedictions of global field models. A reinterpretation of the supporting age controls for these datasets is presented in chapter 6.
Of direct relevance to this study are the results of a palaeomagnetic study on seven lava flows sampled of the Iwikau member, published by Tanaka et al. (1997). The flows were sampled about 1 km north-west of our Delta Corner site and their mean direction (Dec = 9.9°, Inc = -54.2°, α95 = 4.6°) has a similar declination but is more than 10° shallower than our DC flow mean.
Figure 3-7: Comparison of the new palaeomagnetic data (TgVC) with a continuous record from Lake Mavora (black line, with shaded 95% confidence envelope) (Turner et al., 2015b). Our new data is displayed in red. The Lake Mavora record has been relocated to 39.2°S, 175.54°E using a VGP transformation. The uncertainties shown for both the Lake Mavora data and for the discrete datasets have been calculated following the Fisherian formulae for 95% confidence in the mean direction: ∆Inc = α95, ∆Dec = α95/cos(Inc). Not displayed are the palaeomagnetic results from WPC, RT01 and RT02-03 for reasons discussed in the text. We also display the geocentric axial dipole field direction for latitude 39°S (Dec = 0°, Inc = -58°), green dashed line.

Palaeomagnetic refinement of the age-control on flows younger than 11 kyrs
The current, independent age control of the lava flows studied is subject to uncertainties in the range of thousands of years. This is due to the relatively low potassium content of the Ruapehu andesites that were 40 Ar/ 39 Ar dated, and their low radiogenic 40 Ar yields, and to the long time intervals between the tephra marker beds used to bracket the ages of other flows. The availability of the Lake Mavora PSV record covering the past 11, 500 years, together with the high quality of flow mean palaeomagnetic directions, allows us to refine the existing geochronological age control. To obtain an objective correlation, we have employed the Matlab-based archaeomagnetic dating tool of Pavón-Carrasco et al. (2011).
The software package uses a Bayesian statistical approach (Lanos, 2004) to assign age ranges within which an undated discrete palaeomagnetic record with normally distributed uncertainty correlates with a dated master record, to a predefined level of confidence. Probability density functions (PDFs) are calculated separately for declination, inclination and, if the data is available, for intensity. The individual PDFs are then combined into a single PDF. The software package includes a number of global datasets, regional field models and PSV master curves for the northern hemisphere, and also allows the user to import new PSV curves. Here we import the Lake Mavora PSV record (Turner et al., 2015b). We constrain the calculated ages to lie within the ranges given in Table 1 and obtain results to a 95% level of confidence.
Central Crater flow (CC) has a palaeomagnetic direction (Dec = 20.8°, Inc = -67.4°, α95 = 7.4°) close to the present day field (Dec = 21.1°, Inc = -64.5°). It is considered young, from its unweathered appearance and the absence on its surface of pumice from the 232 ± 10 AD Taupo eruption. Palaeomagnetic matching yields an age within the past 500 years. Given that no lava-producing eruptions have been observed from Red Crater in the past century, we assign an age of 300 ± 200 yrs BP to the CC flow.
The very young 40 Ar/ 39 Ar age of the Whangaehu (WE) lava flow (0.2 ± 2.2 ka) should be treated with caution as it is very close to the limit of the technique. The WE palaeomagnetic direction (Dec = 359.5°, Inc = -60.6°, α95 = 3.3°) is northerly and shallower than CC, suggesting an older age. Using 2400 yrs BP as a limit, the combined declination and inclination PDFs define two intervals: 2400 -2000 yrs BP and 700-300 yrs BP (Figure 3-8). Given the extent of weathering and the eroded appearance of the flow (Figure 3-2b), we consider the older interval more likely.
The radiometric ages of the three Iwikau member flows, Taranaki Falls (8.8 ± 2.8 ka), Delta Corner (6.0 ± 2.4 ka) and Bruce Road (ca. 5-10 ka), are statistically indistinguishable. However their mean palaeomagnetic directions differ significantly ( Figure 3-6b), indicating that substantial time probably elapsed between their eruptions. Although the spatial distribution suggests that all flows were emplaced during a relatively short but productive period of effusive activity, small geochemical differences, particularly a higher SiO2 and lower MgO of Taranaki Falls flow (TF) (Conway et al., 2016) support this interpretation. As discussed previously, all three flows and a number of other Iwikau flows sampled by Tanaka et al. (1997) have easterly declinations, and inclinations ranging from -49° to -68°. Between 9000 and 8000 yrs BP, the Lake Mavora record describes a rapid steepening of inclination accompanied by a declination swing from east to west. This is followed by a long period with only low amplitude secular variation from 8000 to 5000 yrs BP. The easterly declinations of our data from the Iwikau member, and the large differences in their inclinations, suggest that they may have been emplaced around the time of these rapid field changes.
The declination of the Bruce Road (RP) flow (Dec = 17.8°, Inc = -60.8°, α95 = 4.0°) is more easterly than any directions in the Lake Mavora record in the assumed age range of 5 to 10 ka. The easterly direction is however consistent with the generally easterly directions obtained from Iwikau flows by Tanaka et al. (1997). While an accurate correlation cannot be made from the declination record, the steep inclination of the RP record suggests that the flow was emplaced after the swing to steeper inclination, and therefore is younger than 9000 yrs BP (probably 8400-8700 yrs BP).
The differences in the site mean directions discussed in section 3.7.3 make it difficult to obtain a precise palaeomagnetic date. However the overall steep inclination suggests ages either at the lower or upper end of the 40 Ar/ 39 Ar range (6.0 ± 2.4 ka).
Unless the Iwikau member flows were erupted over a period of more than 5000 years, then the older of these, ca. 8000 yrs BP, would seem the more probable age of the DC flow.
The Saddle Cone (SC) flows were erupted from a small parasitic cone on the northern flank of Mt Ruapehu. While no direct dating of these flows has been done, Nakagawa et al. (1998) report that the Saddle Cone flows are bracketed by the ca 10 ka Pahoka-Mangamate (PM) pyroclastics and the ca 5 ka Papakai tephra, and are possibly late-stage products of the PM eruption sequence with which they share geochemical similarities. The palaeomagnetic direction for site SC02 (Dec = 358.7°, Inc = -54.9°, α95 = 3.0°) is close to the GAD direction; it falls between the directions obtained from Taranaki Falls (TF) and Delta Corner (DC) lava flows, and is compatible with the Lake Mavora record between 9800 and 8600 yrs BP or between 6100 and 5000 yrs BP. The age estimate obtained from the more easterly direction from site SC01.1 falls into the range of the age brackets provided by SC02. If, as suggested above, the SC lavas do correlate with the end stage of the PM pyroclastic sequence, the older of these intervals would be more likely.
Mangaturuturu flow (MT) has an 40 Ar/ 39 Ar age of 11.9 ± 2.2 ka, overlapping the oldest part of the Lake Mavora record. Palaeomagnetic matching yields a possible age of 10,800 -10,500 yrs BP. An older age must also be considered possible.
Neither the Lake Mavora PSV record, nor any of the current global field models extends beyond 11,500 yrs BP, precluding the palaeomagnetic dating of the older flows studied, including the Whakapapaiti (WPA-C) and Skyline Ridge (GR) flows.
We also refrain from suggesting palaeomagnetic ages for the Rangataua flows. As discussed in section 3.7.10, current age estimates for this voluminous sequence of flows are based only on field relations and are contradictory; the latest field observations suggest that these flows pre-date the Lake Mavora record (Conway et al., 2016).  . a) PSV record for declination and inclination from Lake Mavora and 95% confidence envelope, for the age range provided by the 40 Ar/ 39 Ar eruption ages, with the palaeomagnetic directions obtained in this study (blue line) with corresponding 95% confidence limits (green). Note that for the purpose of palaeomagnetic dating the volcanic records are migrated to the sampling location of the Mavora record (168.2°E, 45.3°S). b) Corresponding probability density functions (PDF, normalized), the green line displays the 95% confidence limit. c) combined probability density functions.

Introduction
The thermoremanent magnetisation (TRM) acquired by lava flows during cooling provides a unique opportunity of obtaining instantaneous and absolute records of geomagnetic palaeointensity beyond times of direct observation. Intensity records give estimates of variations in Earth's dipole moment throughout time and thus yield important information about the dynamics within Earth's core (e.g. Olson, 2002). In order to obtain a reliable record of the global dipole moment, well distributed and unbiased datasets are required. Global databases are disproportionately populated by absolute palaeointensity data from the northern hemisphere (e.g. Donadini et al., 2006;Knudsen et al., 2008). For instance, to date only three studies (Tanaka et al., 1994;1997;2009) have provided absolute palaeointensity estimates for the Holocene time period in New Zealand, yielding a total of seven estimates, some of which are based on four or less successful results. Absolute palaeointensity studies in this region have been hampered by a lack of accurate and precise age control on the volcanic rocks and the stringent requirements for ideal sample material.
The Thellier method is the original palaeointensity method (Thellier and Thellier, 1959) and its modifications are to-date the most reliable technique of estimating absolute palaeointensities from natural TRM bearing samples . It is based on the assumption of a linear relationship between a natural TRM and a laboratory induced TRM. In practice the NRM is replaced stepwise by a laboratory induced partial thermoremanent magnetisation (pTRM) while exposing the samples to a higher temperature with each step. For the linear relation to be valid, the primary remanence carrier should be of single domain (SD) size (e.g. Levi, 1977;Riisager and Riisager, 2001) and of a thermally stable composition, so that the ability to obtain a TRM during an experiment does not change (chapter 2). Neither of these requirements is easy to satisfy. Firstly, the grain sizes found throughout subaerial lava flows are usually within the pseudo-single domain (PSD) or multidomain (MD) grain range (Dunlop and Ӧzdemir, 1997). Secondly the main remanence carriers in intermediate composition lava flows often belong to the Fe-Tioxide ternary system, which includes titanomagnetites and their oxidation products.
Within this system equilibrium temperatures and oxygen fugacities control the composition of co-existing solid-solutions of the spinel (magnetite (mt)ulvospinel (ulv)) and orthorhombic (hematite (hem)-Ilmenite (Ilm)) phases (e.g. Sauerzapf et al., 2008). Rapid cooling or quenching of lava often results in the formation of solid solutions that are metastable at low temperature, and that evolve towards equilibrium composition when reheated during palaeointensity or thermomagnetic experiments. Such alteration affects the samples ability to carry TRM and can therefore result in significant errors in the palaeointensity results.
In this chapter we present the first comprehensive study of palaeointensity in New Zealand on lava flows from the Tongariro Volcanic Centre (TgVC), central North Island. Lava flows were sampled through their vertical thickness and over long lateral distances, targeting samples with a variety of rock magnetic properties. Ferrimagnetic grains in the sampled flows are titanomagnetites in close association with titanium rich titanohematites (chapter 3). Therefore, we are also able to address the effect different oxidation states of these remanence carriers on the palaeointensity results and success rates. In addition to the traditional thermal (Thellier) paleointensity experiments we also used the microwave method, which has the potential to reduce unwanted thermal alteration during the paleointensity experiments (e.g. Hill and Shaw, 2000).
Furthermore in both thermal and microwave intensity experiments we applied the IZZI (Yu et al., 2004) and Coe type (Coe, 1967) protocols. This study is a continuation of the directional PSV study of chapter 3.

Samples
The PSV study described in chapter 3 involved a detailed sampling campaign of 12 andesitic lavas from the Tongariro Volcanic Centre (TgVC). The sampling campaign was carried out in conjunction with a geological mapping programme on Ruapehu Volcano (Townsend et al., in prep), located on the southern end of the TgVC (Figure   4-1). Palaeointensity experiments were conducted on 10 flows, for which reliable palaeomagnetic records were previously obtained. The geology, sampling details, age controls and the palaeomagnetic directions for the selected flows are described in detail in chapter 3 and summarized in Table 4-1.
Where possible lava flows were sampled in several sites and through the flow interior and over large lateral distances. This enabled us to obtain samples with a wide variety of rock magnetic properties and to monitor any terrain-induced errors. Collection of drilled samples was often hampered by the hardness of the andesites. Thus additional hand samples were collected. These were usually removed from the outcrop after marking the strike and dip on an exposed surface, and later drilled with a vertical drillpress in the laboratory. Flat surfaces were not easy to find on the brecciated surface of a'a flow-tops and some hand samples were either poorly oriented or un-oriented.
While these samples were not used for the demagnetisation experiments of chapter 3, they still proved suitable for palaeointensity analysis, where sample orientation is not crucial, as long as a stable remanence direction is known from other, oriented samples (chapter 3).   Ridge forming andesitic flow within Whakapapa ski field. Four sites were sampled throughout the flows thickness and across a lateral distance of ~ 200 m (zones A-C).
A steep river gorge exposes a massive section through the andesitic flow.
Here we present only the results from one site sampled from zone C (site RP02).
Valley bottom flow, exposed along Wairere streambed. Three sites were sampled over a distance of ca. 2 km from the out-washed flow interior (probably zone C).
Andesitic flow, within a sequence of flows, exposed within a steep bluff. Hand samples and core samples were taken from the platey flow interior and directly surrounding rock (zone B).
Ridge forming flow above a sequence of waterfalls. Cored samples were taken from three sites over a distance of ca. 500 m, and additional hand samples from the platey interior (site WPA01) (zones B, C).
Very exposed ridge top flow. We sampled three sites over a distance of ~ 1.5 km. Hand samples were taken from the platey flow centre (zones B, C).
Suite of andesitic lava flows covering an area of ca. 40 km 2 The flow interior is exposed in a small number stream cuts only, which were sampled at sites RT01-RT03. Hand samples were taken from the blocky flow top surface (Sites RT04 and RT05) (zone D).
(8400-8700) < 9000  40 Ar/ 39 Ar dates (Conway et al., 2016), b) tephrochronology (Hobden, 1997;Lowe et al., 2013;Topping, 1974), c) morphology. Pmag age: Refined palaeomagnetic emplacement ages and directions, discussed in chapter 3. Lat, Long are the latitude and longitude of the sampling location, respectively, N the number of sites. Pmag direction: Palaeomagnetic direction. Refer to chapter 3 for a discussion of the statistical procedures applied. Note that for Rangatau flows we separate between two units and directions from site RT01 and sites RT02 & RT03. Refer to section 4.4 for an explanation of the different zones throughout individual lava flows.

Magnetic mineralogy Introduction
Detailed rock magnetic experiments were carried out on samples from each site and supplemented by petrographic imaging. Complementary backscattered electron imaging (BSE) and quantitative analyses of the titanium-oxides were obtained on a selected number of samples (e.g. samples from WE, DC, RP, TF, WPA, WPB flows).
A summary of the rock magnetic findings on lava flows from the TgVC was provided in chapter 3 to the extent needed to interpret the thermal demagnetisation data and to identify the primary carriers of the magnetic remanence. As discussed earlier in this chapter, Thellier-type palaeointensity methods are strongly dependent on the rock magnetic characteristics of the sample material (e.g. Biggin et al., 2007). In this chapter, the rock magnetic behaviour is therefore treated in greater detail, and together with the mineral-optical analyses. Common observations made between individual flows are discussed in the following sections. More detailed descriptions of the rock magnetic properties of every flow are provided together with the palaeointensity results in section 4.6. Data summaries from petrographic imaging and a list of all rock magnetic experiments are available in the appendices.
In chapter 3 we discussed a dependence of the rock magnetic behaviour within individual zones throughout the thickness of the andesitic flows around Mt Ruapehu and we will refer to it in the following analysis. Zone A corresponds to the lowermost part of the flow and is usually not exposed. Zone B corresponds to the flow centre and the lava is usually strongly fractured into curvilinear plates. Zone C is a massive blocky zone that is located between zone B and the auto-brecciated flow top (zone D) ( Figure   4-9b). For details please refer to chapter 3, section 3.4.2 (Whangaehu flow). The Fe 2+/ Fe 3+ cation distribution was calculated from the Fe analysis for identified spinel and rhombohedral phases based on ideal stoichiometry. The coefficients Xusp (ulvospinel) and Xilm (ilmenite) were calculated following Stormer (1983). In this calculation Xusp quantifies the number of Fe 3+ cations replaced by Fe 2+ and Ti 4+ and other cations within the spinel solid solution Fe3-xTixO4, and Xilm is the corresponding mixing coefficient in the rhombohedral hematite-ilmenite solid solution Fe2-xTixO3.

Petrographic description (microscopy and BSE imaging)
The   (Figure 4-3c). Progressive oxidation to a higher state would involve the formation of titanohematite-rutile intergrowth structures. In this study however, the highest oxidation state observed corresponds to index 3.
Later in this chapter (e.g. section 4.4.5) we describe differences in the rock magnetic properties in relation to their vertical position within the flows. At this state it is noteworthy that although the relative abundance and size of micro-lamellae in larger grains and the distribution of submicron sized grains within the groundmass vary strongly from flow to flow, we did not identify significant differences between different sites sampled on individual flows within the visible range (≥ 1 μm).

Quantitative analysis (WDX)
The quantitative wavelength dispersive x-ray (WDX) analysis has a maximum resolution of 5 μm and was therefore carried out on phenocrysts only. While co-

Rock and thermomagnetic behaviour
The petrographic and quantitative analysis described above suggest that the magnetic minerals in samples from all sites and flows consist primarily of titanomagnetites or titanohematites with low to moderate oxidation states (indices 1 to 3 after Watkins and Haggerty, 1968).
The rock and thermomagnetic properties agree well with this observation.
Isothermal remanent magnetisation (IRM) and backfield coercivity of remanence curves from all samples strongly resemble each other. The IRM curves saturate at applied fields lower than 250 mT and require backfields (coercivities) between 15 and 30 mT (Bcr) to demagnetise samples thereafter, as would be expected from these relatively soft remanence carriers (Figure 4-6).
The magnetic coercivities for most samples range from 5-15 mT. Only samples from Central Crater (CC) and Bruce Road (RP) flow yield higher coercivities up to 30 mT ( Figure 4-7). The ratios between saturation remanent magnetisation (Mrs) and saturation magnetisation (Ms), coercivity of remanence (Bcr) and coercivity (Bc) fall into the range of values expected for pseudo-single domain (PSD) grains defined by Day et al. (1977) and Dunlop (2002) (chapter 3, Figure 3-3a). However based on the wide variation of grain sizes observed during mineral-optical analyses we suggest that these ratios may be representative for an average between different coercivity populations. As discussed in chapter 3, the thermomagnetic curves also vary strongly from site to site. The Curie temperatures range from 150 to 600°C, the lower end of the range agrees roughly with the Tc expected for the composition measured during WDX analysis, while the higher Curie temperature (Tc > 550°C) is higher than the Tc expected from any of the compositions measured (Table 4-3) (Lattard et al., 2006), suggesting the presence of low titanium-titanomagnetite phases with effective magnetic grain sizes that are smaller than the minimum resolution of the electron microprobe (< 5 μm). In chapter 3 we summarized the different thermomagnetic curves into five types, based on the Curie temperature distribution and the reversibility of heating and cooling curves (Table 4- We suggest that these differences are the result of differing cooling rates throughout the andesitic flows sampled. The cooling, degassing and crystallisation process during and after emplacement of the lava flows may have caused a shift of the equilibrium composition within the spinel-orthorhombic system described by the Fe-Ti oxides (e.g. Sauerzapf et al., 2008). While quenching of the lava on the flow top preserved titanomagnetites close to their primary state (e.g. oxidation index 1 or 2, after Watkins and Haggerty, 1986), cooling within the flow centre may have lasted up to weeks (Sigurdsson, 2000), which would have allowed the titanomagnetites to undergo subsolidus exsolution and oxidation. As discussed in section 4.4.3, these within flow differences were not directly visible during petrographic imaging and we thus suggest that oxidation after or during emplacement of the lavas resulted in the formation of sub-micron scale lamellae or additional growth of sub-micron sized grains in the rock matrix, which are optically unresolvable.       Lattard et al. (2006) in comparison to the Tc's obtained from Ms vs.T and χ vs.T curves on sister samples. The Xusp component was calculated from the microprobe results following the steps outlined in Stormer (1983). Tc's were estimated using the second derivative approach from the Ms vs.T curves and hand-picked from the onset of linearity in 1/χ from the χ vs.T curves (Petrovskӯ and Kapicka, 2006  The heating curves show two or more phases with T c 's between 250 and 320°C (P2) and between 320 and 450°C (P3). The Tc's change after heating, the Tc shift is dependent on the peak temperature during an experiment.
Characterised by a single phase of high Tc (P4), the heating and cooling curves are near reversible.

Palaeointensity methods
Palaeointensity experiments were carried out on a broad range of sites with a variety of rock magnetic properties, using both thermal and microwave palaeointensity procedures and applying Coe (Coe, 1967) and IZZI-type (Tauxe and Staudigel, 2004) protocols, as outlined in chapter 2. Adjusting the applied field near the actual palaeointensity can reduce the effect of experimental errors (Tanaka and Kono, 1984).
We thus used a laboratory field of 30 μT during the first suite of experiments, and a field of 50 μT during later experiments, which spans the limits of intensity expected of the palaeomagnetic field at this latitude.

Thermal palaeointensity method
Up to 10 standard sized specimens per site were prepared for palaeointensity experimentation. Specimens were grouped according to their blocking temperature distribution and intensity experiments designed correspondingly. First palaeointensity experiments were carried out using the IZZI protocol and additional Coe-type experiments were conducted on specimens from sites that showed promising results during the first suite of experiments. All thermal experiments were accompanied by low field susceptibility checks using a Bartington MS2 in order to monitor mineral alteration. Thermal palaeointensity experiments were carried out using either a Magnetic Measurements Ltd. slow or fast cooling oven for thermal de-and remagnetization and measured using an AGICO JR6 Magnetometer at the University of Liverpool and an AGICO JR6A at Victoria University of Wellington. Prior to commencing the in-field steps, a profile of the total field strength along the oven was measured using a fluxgate magnetometer. The samples were placed within the most uniform (field variation max. 1%) area within the centre of the oven and at each step held for 30 minutes at their respective temperature. The cooling times from temperatures between 300 and 600°C to near room temperature were approximately 30 minutes in the fast cooling, and up to two hours in the slow cooling oven. Sample orientation in the laboratory field can affect the outcome of pTRM tail checks and the palaeointensity result, if the samples suffer from strong remanence anisotropy. For instance, application of the field applied at a high angle to the NRM direction during infield steps enhances the effect of pTRM tails (Yu et al., 2004), while application of the laboratory field parallel to the NRM direction minimizes effects of magnetic anisotropy (e.g. . Because the laboratory set-up used for the thermal experiments only allowed specimens to be placed in the oven with the laboratory field applied along the core-axis, the specimens were placed in the oven so that, alternating between specimens, the field was applied either parallel or antiparallel to the z-axis.
Alternating between specimens in this way we aimed to record either effect.
Additionally, some palaeointensity experiments were aborted when the onset of thermal alteration became apparent from the susceptibility measurements.

Microwave palaeointensity method
Microwave palaeointensity experiments were carried out using the 14.5 GhZ system at the University of Liverpool, described in detail in chapter 2. The technique uses ferromagnetic resonance (FMR) to excite the magnetic moments of the grains rather than heating the bulk sample but otherwise follows similar experimental procedures as the thermal experiments. For comparison, microwave experiments were carried out using similar protocols and, where possible, on specimens from the same samples of those used for thermal experiments. Sample size and positioning within the cavity varied slightly and the exact frequency (range 14-14.5 GhZ) was fine-tuned prior to each experiment in order to achieve maximum absorption rates. The parameter affecting the demagnetisation level is the total energy absorbed by a sample, which is a function of the power and exposure time of microwaves. Slight variation of the energy absorption can be caused by movement of a sample within the cavity and/or mineral alteration. Repeatability of individual steps is a prerequisite for both Coe and IZZI type protocols and we rejected steps where the absorption varied by more than 10% from the first step at a given power level. The applied power was usually increased from a starting value of 10 W in steps of 2-5 W, using an exposure time of 5 s. If samples required further treatment after reaching 30 W, we usually increased the exposure time. Previous microwave studies (e.g. Stark, 2011) have shown that this approach minimizes dielectric heating of the samples.
In contrast to the thermal experiments, microwave experiments were conducted on one specimen at a time, which enabled us to adjust the power steps according to the behaviour of the individual specimens. Prior to each intensity experiment we conducted a rapid demagnetisation experiment on each sample to design suitable power steps and eliminate samples that showed strong overprinting.

Selection criteria and calculation of flow mean palaeointensities
Palaeointensity selection criteria are applied to data in order to ensure that only high quality and reliable palaeointensity results are accepted. All palaeointensity data in this study were analysed using ThellierTool v4.22 (Leonhardt et al., 2004a) which outputs the most commonly used statistical parameters that quantify the quality of the slope fit within the Arai-plot as well as pTRM-and tail checks.
A range of different statistical parameters, combinations of these and values have been proposed in the past (e.g. Biggin et al., 2007;Kissel and Laj, 2004;Leonhardt et al., 2004a) and modified in later studies. In a recent effort to produce objective sets of criteria that effectively maximize the number of accurate results accepted and erroneous results rejected, Paterson et al. (2014) tested the most popular sets of selection criteria. Amongst the criteria sets examined in their study were ThellierTool software's internal default criteria TTA and TTB (Leonhardt et al., 2004a), where TTB is the less stringent. Paterson et al. (2014) found both criteria sets effective but suggested slight modifications, which have been adopted in this thesis and are referred to as TTA* and TTB* and are summarized below: 1) The zero-field remanence measurement should show straight line decay to the origin, the anchored fit (α) describe an angle no larger than 15° with the free floating fit and yield a maximum angular deviation (MADanc) smaller than 6°o r 15° to pass TTA*and TTB*, respectively.
2) Palaeointensity estimates should be calculated from at least 5 data points and include at least 35% of the total NRM (f ≥ 0.35).
3) The quality of the slope fit in the Arai-diagram (β) is quantified using the standard error of the slope normalized by the absolute value of the slope. β ≤ 0.1 to pass TTA* and β ≤ 0.15 to pass TTB*. Further we calculated the overall quality factor q=fgβ -1 where g is the gap between any two data points. q ≥ 5 to pass TTA* and q ≥ 0 to pass TTB* 4) The maximum absolute difference produced by a pTRM check is calculated by vector subtraction and divided by the total TRM, dCK ≤ 7% and dCK ≤ 9% to pass TTA* and TTB*, respectively. The cumulative pTRM check failure (dpal) is quantified from the difference between the slope of an uncorrected palaeointensity estimate and the slope, corrected by the added effect of the alteration (see Valet et al., 1996), normalized by the uncorrected slope: dpal = 100 − * where b is the uncorrected and b* the corrected slope.
Flow mean palaeointensities were averaged from all successful results, unless we identified systematic inter-site differences (e.g. see section 4.6.2, Whangaehu flow).
Sample orientation is not crucial, so in contrast to the analysis on directional results in chapter 3 we did not calculate individual sample or site means. b dpal ≤ 10% (TTA*), dpal ≤ 18% (TTB*).

Palaeointensity results
Successful palaeointensity results were obtained on samples from six flows. The behaviour and success rates of individual specimens during the palaeointensity experiments varied from site to site in accordance with the differences observed between the rock magnetic properties. The palaeointensity and rock magnetic results One intriguing microwave result was obtained from site CC02 (CC12A2, Coe protocol). Despite obeying ideal linearity and passing both tail and pTRM checks, the Arai plot for this sample yields approximately 20% higher palaeointensity than the results obtained from sister samples using both the thermal (e.g. CC13B, CC13C) and microwave (e.g. CC12A1) palaeointensity method (Figure 4-10b, c). The results show that even the highest quality results can yield significant different palaeointensities.
Our best mean palaeointensity results, calculated by averaging all results that pass at least the selection criteria TTB*, is 70.6 ± 4.1 μT (N = 6).
Microwave palaeointensity experiments were conducted on a broad range of sample material from throughout the flow, while thermal experiments were focussed on hand samples taken from the flow top (zone D) and flow centre (zone B).
All palaeointensity experiments were carried out with a laboratory field of 50 µT.
The rock magnetic differences reflect on the palaeointensity results from various sites.
We expected higher success rates from the flow centre (zone B) samples, for which the higher coercivities suggested higher resilience against viscous remanences and the thermomagnetic curves indicated thermal stability. However, the contrary was the case. Most thermal and microwave palaeointensity results on zone B samples were unsuccessful, evident from strong zigzagging in the Arai plot about a median line during IZZI-type experiments, and associated with a concave shape and pTRM check failure (Figure 4-11i). Zones A and C and D gave higher success rates, where six out of eight experiments pass either TTA* or TTB* (Figure 4-11e, f). However, the palaeointensity estimate calculated from a flow top sample differs significantly from the palaeointensity results from all flow interior samples (zones A, B, C). The palaeointensity estimates from successful thermal and microwave experiments from zone D range from 51 to 55 μT (mean: 54.4 ± 1.9 μT (N = 3), Figure 4-11b,c) and are approximately 20% lower than the estimates made on samples from the flow interior (mean: 63.6 ± 3.0 μT (N = 5)). There is experimental evidence that the correct palaeointensity result lies somewhere between the highest and the lowest palaeointensity result obtained (e.g. Biggin et al., 2007). We thus present our best estimate as the mean of the two averages (flow interior zones A-C, and the zone D average), and provide an uncertainty that encompasses the mean and standard deviation of the two extremes, giving: 59.6 ± 7.0 μT (N = 8).

Delta Corner flow (DC)
The rock magnetic behaviour of samples throughout the thickness of Delta Corner flow resembles those described for Whangaehu flow. All samples were taken from zones A-C, hysteresis loops exhibit dimensions typical for PSD grain sizes (Bc ~ 10 mT) and the thermomagnetic curves exhibit type 3 or 5 behaviour. Six thermal and four microwave experiments were carried out using either the IZZI or Coe type   (Figure 4-13b,d). This is consistent with the thermomagnetic behaviour and we thus attribute the two demagnetisation stages to the two phases, P2 and P4.
Each phase carries approximately 50% of the total remanence. In both the thermal and microwave applications, the pTRM checks fail at the saddle point between the demagnetisation of P2 and P4 (Figure 4-13b, c). During heating experiments, this point also marks the onset temperature of thermo-chemical alteration (T~350°C), evident also in the accompanying susceptibility checks. All palaeointensity results yield a stable relation between NRM and pTRM within the Tb range NRM to 350°C and 6 out of 11 specimens pass the most stringent criteria set TTA* (Table 4-4). Our best mean intensity, including all accepted data averages to 51.2 ± 3.5 μT (N = 9). All samples from Taranaki Falls (TF) flow are from zones B or C and yield typical type 3 or 4 thermomagnetic curves such as observed at Whangaehu and Bruce Road flows (e.g Figure 4-14a). In contrast, the hysteresis curves are narrower (Bc < 8 mT).
Thermal and a small number of microwave palaeointensity experiments were carried out, using a laboratory field of 30 μT. The overall success rate is with 8 out of 15 results that meet either TTA* or TTB* criteria lower, but all successful palaeointensity results strongly resemble those of Bruce Road flow: zero-field remanence measurements show two distinctive Tb ranges (not displayed in Figure   4-14). The lower Tb range carries between 45 and 53% of the total remanence. Thermal alteration results in failure of the experiment at intermediate demagnetisation/ treatment levels ( Figure 4-14 b,c). The rejected palaeointensity results are characterised by strong zigzagging of the remanence intensity during IZZI-type experiments and the zero-field direction deviates into the direction of the laboratory field used during infield steps (Figure 4-14d,e). Such a bias results when Tub > Tb, and the sample retains a small remanence in the field direction. This can be caused either by a lack of symmetry in domain wall movement between zero-and infield steps, and therefore be indicative for the presence of MD grains (e.g. Levi, 1977) or it may result when a low Tb phase alters into a phase of higher blocking temperature. The presence of MD grains is justified by the low coercivities measured on Taranaki Falls (TF) samples.
Palaeointensity values from experiments that pass the most stringent selection criteria (TTA*) are within 3 μT of each other and fall within our best mean and its uncertainty: 37.0 ± 5.7 μT (N= 8), which was calculated including the lower quality results, passing TTB* criteria.

Mangaturuturu flow (MT)
Although Mangaturuturu flow was sampled at one site within the platey flow centre (zone B) only, the thermomagnetic behaviour varied considerably between individual samples. All Ms vs. T curves are irreversible and can be roughly classified as type 3 or 4 (e.g. Figure 4-15a, Figure 4-16a, c). To accommodate the difference in the rock magnetic behaviour, microwave and thermal palaeointensity experiments were carried out on a broad range of sample material using a 50 μT laboratory field. Most samples show viscous overprints that justify exclusion of low power or temperature steps.
Experimental failure was either caused by non-linearity of the relation between NRM and pTRM displayed in the Arai plot (e.g. Figure 4-16c), or a deflection of the remanence direction from the origin (e.g. Figure 4-16b). Five out of nine microwave and four out of eight thermal palaeointensity experiments pass either set of selection criteria (Table 4-   Microwave palaeointensity experiments were carried out on samples from sites RT01, RT05 and RT06, and thermal Thellier type experiments on hand samples from sites RT05 and RT06. Most samples carry a viscous overprint that is removed after the first treatment step. Because the magnetic remanence is primarily carried within the Tb range < 320°, all palaeointensity experiments are completed prior to the usual onset of thermo-chemical alteration (Figure 4-17b ,c). However, some specimens were affected by strong viscous remanences, resulting in low quality results that did not meet the selection criteria. In chapter 3, we suggested that despite the rock magnetic similarities, site RT01 belonged to a lava flow that was emplaced separately from a flow containing sites RT02 and RT03. For the latter sites, we did not obtain successful palaeointensity data. The palaeointensity results from sites RT01, RT05, RT06 are statistically indistinguishable, allowing to calculate an average of 32.0 ± 5.1 μT (N=8).  Table 4-4: Accepted palaeointensity results from the TgVC. A summary of all palaeointensity experiments carried out is available in the appendices. The top two rows define the criteria sets used, which are based on Leonhardt's (2004a) criteria sets TTA and TTB, but include modifications suggested by Paterson et al. (2014). A small number of data are included that do not pass the stringent criteria sets, but for which failure can be attributed to known and acceptable sources (refer to notes below table). Note: Spec. ID corresponds to the specimen ID (shortened), Full ID to the specimen ID, Type: M = microwave palaeointensity, T = thermal palaeointensity, Prot. is the experimental protocol used, Interv. is the temperature interval/microwave steps selected, n,f, β, q, MAD dCK, dpal, dTR, dt* are statistical parameters that were calculated following the standardized palaeointensity definitions (Paterson et al., 2014), whereas MAD corresponds to the maximum angular deviation when anchored to the origin. Int is the palaeointensity, and the standard error σ. Sel. are the selection criteria passed. A = TTA*, B = TTB* as discussed in the text. Averaging: N/n = number of successful specimens included in average/ total number of experiments carried out. * pTRM checks did not meet the selection criteria due to the inclusion of high temperature/power steps. The additional data were only included when the slope of the high temperature/power steps is consistent with the one of lower temperature/power steps. ** pTRM check failure associated with inconsistencies in the energy absorbed during microwave treatment. *** Our best mean palaeointensity for Whangaehu flow was calculated from the means of flow interior (zones A, B, C) and flow top (zone B) samples and encompasses their respective confidence limits

Palaeointensity success rates and data quality
The results of all palaeointensity experiments and selection criteria sets passed are summarized in Table 4 failure usually occurred at a temperature > 350°C, which corresponds to the onset temperature of thermo-chemical alteration known from the repeated measurement of χ vs. T curves, with an increase in the temperature following each heating and cooling cycle (section 4.4.2), and also from accompanying susceptibility checks. However, the temperature in the microwave cavity is not expected to exceed 250°C (Hill and Shaw, 2000;Suttie et al., 2010) and it is therefore questionable what process leads to the pTRM check failure in these experiments.
The palaeointensity estimates made using thermal and microwave experiments by applying IZZI and Coe-type protocols compare well and no systematic offset was identified between the results from microwave or thermal experiments (Figure 4-18a).
Palaeointensity estimates were made from specimens sampled over wide lateral distances (e.g. TF) and including a wide variety of rock and thermomagnetic properties (e.g. CC, MT). The standard deviation about the flow mean intensities, calculated from all successful results including both microwave and thermal palaeointensity results is usually 10% or less of the mean (TF, MT flows are exceptions) and the individual specimen intensities do not deviate by more than 20% of their mean (Figure 4-18). The range of accepted values can be further reduced by the application of more stringent selection criteria (e.g. TTA*). However, it has frequently been discussed that too stringent data selection can result in removal of useful data (Biggin et al., 2007;Paterson et al., 2014) and, as shown in section 4.6.1 on specimens from Central Crater flow (CC), even the highest quality results may be erroneous. Removal of lower quality data may therefore place unduly high confidence on individual results and bias the resulting average.

Origins of data scatter and experimental failure
Systematic differences between the palaeointensity results obtained from units of the same age, potential sources for data scatter and failure in the palaeointensity experiments have been described widely (e.g. Biggin et al., 2007;De Groot et al., 2014;Yamamoto et al., 2003). and P2, which have Tc's below 320°C. We assume that magnetic mineral alteration was caused by a shift in the equilibrium composition of the titanomagnetites during re-heating.
If formed below the Curie temperature, for example during progressive exsolution, ferro-/ferrimagnetic grains carry a thermo-chemical remanence (TCRM) rather than a TRM (Draeger et al., 2006), resulting in non-linearity in the Arai plots (e.g. McClelland, 1996). If present we would expect a TCRM to be carried by titanomagnetites of compositions P3 and P4, which most probably formed from the low-Tc primary phases P1 and P2.
Weathering of a lava under frequent exposure to snow and water can result in the lowtemperature transformation of (titano)magnetites to metastable (titano)maghemites (Tauxe, 2015). . It is also possible that sub-micron sized grains formed within the rock matrix.
Despite the overall higher coercivities observed, samples from the flow centres (zones B, C) often showed pTRM tails, characteristic of Tub > Tb, caused by MD behaviour (e.g. section 4.6.5). We suggest that in mixtures of SD and MD titanomagnetites with high Tc's (Tc > 450°C), low Tb unblocking is dominated by low coercivity unblocking of the MD grains. In contrast, in samples that are dominated by magnetic mineral phases P1 and P2 with Tc's < 320°C, unblocking of the entire coercivity range occurs at low temperature, and the overall MD effect is thus significantly smaller. MD behaviour is also known to result in concave upwards shape of the relation NRM/TRM displayed in the Arai plots (Levi, 1977), which may explain the higher palaeointensities measured from the Tb range < 350°C on flow centre samples, such as found on Whangaehu lava flow.
TRM acquisition is not only a function of temperature but also cooling time (e.g. Biggin et al., 2013;Bowles et al., 2005;Leonhardt et al., 2006) and in theory variation in the cooling rate throughout the thickness of the lava flows may also directly affect the palaeointensity results (e.g. Biggin et al., 2007). In the laboratory we applied two different sets of cooling rates: During thermal experiments, cooling from temperatures between 100 and 600°C lasted between 30 minutes to two hours (see chapter 2), while the maximum treatment interval during microwave experiments was 15s.
Nevertheless, the results from both experiments are similar, which suggest that on these samples cooling rates had only a minor primary effect on the palaeointensity results measured. However we acknowledge that the theoretical foundation for the TMRM acquisition in the microwave is not adequately established. The cooling rate effects may be quantified experimentally (e.g. Biggin et al., 2013) but further experimentation was beyond the scope of this thesis.

Errors associated with averaging
The best estimate palaeointensity for all but one flow was calculated from all successful specimens by averaging. This approach is based on the assumption that the data is distributed normally around the mean and is subject to random uncertainty or noise. In this approach, highest confidence is usually placed on estimates with the lowest standard deviation.
However there is evidence that the noise present in our data is not random. For example, Whangaehu lava flow samples exhibit a bias, which varies with the location in the vertical position of the sampled lava flow. As discussed in section 4.7.2 such a bias is not uncommon (e.g. Biggin et al., 2007) and is probably linked to factors such as MD-type behaviour, cooling rate or other rock magnetic contributions. The implication is that if a palaeointensity estimate is calculated from the results obtained from one location in the thickness of a flow only, the estimate may have a low standard deviation but be inaccurate. In the present study, this applies to Bruce Road flow (RP): with a standard deviation of 6% about the mean and overall high quality (TTA*) results the data appears reliable. However the results were obtained from within one zone of the flow only with little variation in the rock magnetic behaviour and the results may thus be biased to either too high or low palaeointensity results.
Highest confidence should therefore be placed on the palaeointensity results from flows where consistent results were obtained from spatially separate sites and where consistent results were obtained from samples showing differences in the rock magnetic behaviour (e.g. Taranaki Falls flow). Such ideal results are rarely observed.
While sources for data scatter or intensity biases have been discussed widely (e.g. Biggin et al., 2007;De Groot et al., 2014;Yamamoto et al., 2003), Simple averaging as done in this thesis may thus exhibit significantly larger uncertainty than suggested from the standard deviation provided. To improve such results a better understanding how rock magnetic mineralogy and other factors relate to the palaeointensity recorded is needed. Lake Mavora (Turner et al., 2015b). A priori, the sedimentary record provides relative palaeointensity only. To make it comparable to the discrete data we apply the scaling factor suggested by Turner et al. (2015b), which is based on the earliest absolute intensity recordings from New Zealand prior to relocating the data to Whakapapa Village (39.2°S, 175.54°E) using a VGP-transformation (refer to chapter 3 for details).

Comparison with previous datasets from New Zealand
We also display the field predictions from the latest spherical harmonic global field model pfm9k (Nilsson et al., 2014).  (Tanaka et al., 1997) and site RP02 (this study) does not allow a field correlation, however based on the PSV data this suggests that the sites either belong to the same lava flow or to flows that were emplaced during a short time-period.
Our new data also correlates exceptionally well with the scaled Lake Mavora Curve within uncertainties of the palaeomagnetic ages of chapter 3. We opted not to include intensity as a third component in the palaeomagnetic dating, due firstly to the additional uncertainties from the data scatter described above and secondly the scaling of the relative intensity record. However to a first order, the correlation shown in Figure 4-19 strongly supports the palae omagnetic dating results presented. Figure 4-19: Records of palaeomagnetic declination, inclination and intensity from the TgVC. Displayed are the discrete directional records discussed in chapter 3 and the new palaeointensity data for the Holocene time-period presented in this chapter superimposed on the continuous lake sediment record from Mavora (Turner et al., 2015b) and the global field model pfm9k (Nilsson et al., 2014) (green line). Prior the relocation, the relative palaeointensity curve from Lake Mavora was scaled using a constant factor of 64 μT. The discrete data is displayed with our preferred palaeomagnetic ages presented in chapter 3 (red) and the age brackets of the independent constraints (dashed lines).

Conclusions
This chapter presents detailed rock magnetic and palaeointensity studies on the andesitic lava flows of Mt Ruapehu and Tongariro described in chapter 3. New palaeointensity estimates were made on six andesitic lava flows from the Tongariro Volcanic Centre (TgVC) in New Zealand using a combination of microwave and thermal Thellier-type palaeointensity procedures. The rock magnetic properties vary considerably through the thickness of individual flows, and differing success rates and data dispersion were linked to these. The new discrete, absolute palaeointensity data range from 37.0 ± 5.7 μT to 70.6 ± 4.1 μT. These values fall well into the range of palaeosecular variation swings suggested by the scaled relative palaeointensity curve from Lake Mavora (Turner et al., 2015b) and describe a similar overall trend from low to higher palaeointensities throughout the Holocene, as is also suggested by the global field predictions from model pfm9k (Nilsson et al., 2014). Palaeointensity experiments were carried out on four lavas using the Coe-type double heating method. Succesfull palaeointensity estimates, made on three units range from 51.7 ± 3.2 μT to 63.3 ± 1.2 μT. The new palaeomagnetic data generally compares well or is an improvement to the palaeomagnetic results published on the same lavas during previous studies.

Introduction
Accurate, well-resolved records of palaeomagnetic direction and absolute palaeointensity through time are required to understand the dynamics of the outer core of the Earth -the geodynamo. Over the last decade much effort has been invested in creating global data compilations Korte et al., 2005, and following publications), and in setting up global databases such as GEOMAGIA50 (Brown et al., 2015;Donadini et al., 2006;Korhonen et al., 2008), MagIC  and Pint (Biggin et al., 2009;Biggin et al., 2010). The geomagnetic field observed on Earth's surface is traditionally modelled by least squares fitting of the available data to spherical harmonic functions. Such analysis is carried out both from direct observations on centennial (Jackson et al., 2000) and palaeomagnetic data on millennial (Korte et al., 2011;Nilsson et al., 2014) time-scales. However, all models and global databases are strongly affected by the quality and spatial distribution of the data. For instance, the statistical data treatment and the methodologies in palaeomagnetic analysis and in particular for the determination of palaeointensities improve continuously and the quality of the data previously published and that have entered databases is therefore highly variable (e.g. Donadini et al., 2006).  (1994; 2009), however this data included extreme easterly and westerly declinations and very high intensities that are not typical of the PSV of a stable normal polarity field. We finally revise all discrete datasets on Holocene (< 15 ka) volcanic materials and compare the data to regional datasets and global field reconstructions.

Geological Setting
New Zealand's youngest and current volcanic activity is focussed within the Taupo Volcanic Zone (TVZ), an area of high geothermal flux, active extension and voluminous rhyolitic volcanism (Bibby et al., 1995;Wilson et al., 1995) ( Figure 5-1, inset). Some activity is also seen in the monogenetic Auckland Volcanic Field and around Mt Taranaki ( Figure 5-1, inset). The TVZ results from subduction of the Pacific plate beneath the Australian plate along the Hikurangi-Kermadec Trench system; it strikes roughly north-east from the Tongariro Volcanic Centre (TgVC) to White Island in the Bay of Plenty, from where it extends into the oceanic Kermadec-Tonga arc-back-arc system. Over the last ~ 60 ka the silicic eruptive activity has focussed on four major volcanic centres: Okataina, Maroa, Kapenga and Taupo (Wilson et al., 1995).
The Pleistocene eruptive activity of Taupo Volcano was dominated by the 25.4 ± 0.2 ka Oruanui supereruption, which formed much of the present day Taupo caldera (Van Eaton and Wilson, 2012;Vandergoes et al., 2013). At least 28 distinct eruptive episodes have occurred since (Wilson, 1993) and while the early activity was dacidic in composition, at least 10 rhyolitic eruptions have occurred in three distinct periods through the Holocene (Barker et al., 2014). The eruptive stratigraphy is primarily defined from pyroclastic deposits and the vent sites are mostly inferred to be submerged beneath the eastern part of Lake Taupo. Only four associated lava flows or domes are exposed on Lake Taupo's eastern and northern shores (Wilson, 1993).
The most prominent and latest eruption from Taupo Volcano occurred 1.72 ± 0.01 ka ago and produced widespread pyroclastic and airfall deposits, including the Taupo Ignimbrite, a frequently discussed marker bed within the central North Island (Houghton et al., 2010;Lowe et al., 2013).
More recent activity has occurred in the Okataina Volcanic Centre (OVC) -the historic basaltic Tarawera eruption occurred only 130 years ago. Volcanic activity around the OVC is believed to extend back to at least 550 ka (Cole et al., 2010) and its evolution is defined by at least four major phases of caldera collapse, which are associated with widespread pyroclastic and tephra deposits. The youngest (< 61 ka) and presently dominant caldera is Haroharo, which covers an area of approximately 450 km 2 (Nairn, 2002). It is infilled by the products of at least eleven distinctive eruptive episodes during the last 26 kyrs that have formed the massive volcanic dome complexes of Haroharo at the northern end and Tarawera in the south (Figure 5-1b). Each episode included, in short succession, voluminous pyroclastic, plinian and effusive eruptions (Smith et al., 2006), sourced from multiple, simultaneously active vents that are aligned along two linear zones that include the Tarawera and Haroharo Volcanic Complex (Nairn, 2002;Smith et al., 2006). Plinian pyroclastic eruptions dispersed tephra over much of the central and eastern North Island, and these form characteristic marker beds in distal sedimentary sequences, which are often dated by radiocarbon methods. Based on these, a stratigraphic framework has been developed (Froggatt and Lowe, 1990 and revised by Lowe et al., 2013) and adopted for the naming of associated eruptive episodes that formed voluminous vent-proximal pyroclastics deposits, ignimbrites, lava flows and domes. All these deposits have been summarized in the 1:50,000 scale geological map of the Okataina Volcanic Centre (Nairn, 2002). Nairn (2002) showed that the radiocarbon ages of a number of proximal deposits are within 100 to 300 years of Frogatt and Lowe's (1990) tephra ages and thus assigned each eruptive episode with an "accepted" age that was rounded from their stratigraphic and radiocarbon ages. Since the geological map was published, the geological framework has not changed significantly, however many of the tephra ages have been revised .

Samples
Palaeomagnetic field sampling was carried out in October 2012 and August 2014.
Field sampling for the OVC was limited to its northern end, for which sampling permission was given by local by iwi and forestry companies. The sampled sites were selected based on the accessibility in the field and from outcrops that appeared in-situ. Altogether we sampled ten sites from six rhyolitic lava flows or domes from the OVZ that were emplaced during the Whakatane, Mamaku, Rotoma and Waiohau eruptive sequences. We also sampled four sites on a dome structure located at the northern end of Lake Taupo. At each site we sampled cores of 2.5 cm diameter and 10 cm length using a water cooled hand drill. All samples were oriented using a suncompass. The site locations often correspond or are near the location from previous studies, both are displayed in Figure 5-1. The sampling details, geological and age information, and palaeomagnetic directions for all lavas sampled in this study are summarized together with the previously published data Table 5-2. The sampling details are also discussed in detail, flow by flow below.
The tephra ages cited for the individual eruptive sequences correspond to revised and calibrated radiocarbon ages proposed by Lowe et al. (2013). A more thorough discussion for these ages is provided in chapter 6.

Tapahoro flow (TC)
Tapahoro flow is part of a sequence of flows, sourced from the Haroharo Volcanic Complex. The flow margins form steep cliffs above the northern end of Lake Tarawera and its outlet area in the east (Figure 5-1b). Two sites, TC01 and TC02, were sampled, the sampling coordinates suggest that both sites are located within 200 m of the sampling coordinates of site NK02 of Tanaka et al. (2009). The flows are believed to have been emplaced during the 5526 ± 145 cal yrs BP Whakatane eruptive episode.

Haroharo dome (HR)
This lava dome covers an area of roughly 1 x 2 km and rises to a high point ~ 200 m above the Haroharo Volcanic Complex. It is the only unit addressed in this study for which no previous palaeomagnetic data is available and it was sampled at two sites. Site HR01 is located in a streambed on the southern flanks of the dome complex, site HR02 is in a small quarry site near the top of the dome ca. 1 km from site HR01 and was sampled at two sub-sites. It is located on top of much of the Haroharo Volcanic Complex and thus believed to have been emplaced during the Whakatane sequence as well.

Hainini dome (HN)
Hainini dome is a ca. 1.5 km 2 rhyolitic dome at the centre of the Haroharo Volanic Complex. It was emplaced alongside numerous other flows during the 7940 ±257 cal yrs BP Mamaku sequence. The steep northern face of the dome was sampled at two sites, separated by a distance of 300 m. Site HN01 was sampled from a road cut and corresponds to site NK05 of Tanaka et al. (2009). Site HN02 was sampled from a steep bluff ca. 50 m from the closest access point on the forest road.

Waiti flow (WT)
Waiti flow is situated in a densely forested area which provides little exposure of the flow interior. In the 1:50,000 geological map (Nairn, 2002) it was first associated to the Mamaku eruptive episode. Based on recent geochemical findings by Schmith (2006) it is more likely that it was already emplaced during the earlier 9423 ±120 cal yrs BP Rotoma eruptive sequence. The flow was sampled in a small road-cut, approximately 20 m in length, at two sub-sites, WT01.1 and WT01.2.

Rotoma flow (ROT)
Rotoma lava flow is exposed along the southwestern lake shore of Lake Rotoma ( Figure 5-1b). At its highest point it is rises ca. 100 m above lake level and the terrain drops steeply into the lake on its northern end where most of the surface exposure is found. We sampled two sites: Site ROT01 is located on the north-eastern lake shore and was sampled at several sub-sites over a distance of 30 m. Site ROT02 is located near the flow top: sub-sites (ROT02.1, ROT02.2) comprise weathered road-cuts, 50 m apart. The flow was also emplaced during the 9423 ±120 cal yrs BP Rotoma eruptive sequence.

Acacia Heights dome (AH)
Acacia Heights is a ca. 1.5 x 1 km 2 rhyolitic dome, located near Acacia Bay on the northern shore of Lake Taupo. It is one of four lavas associated with Holocene silicic eruptions from Taupo Volcano (Wilson, 1993). Like every other rhyolitic lava in this study its emplacement age is defined by correlation with tephra deposits. It was first associated with the 11,170 ± 115 cal yrs BP Poronui tephra (Froggatt, 1981). However recently Wilson (1993) suggested that it is located on the inferred vent location of a deposit (unit D) that is approximately 20 years younger than the Poronui tephra. We thus infer a tephra age of 11,150 ± 115 cal yrs BP.
The dome was sampled at three sites on the upper surface of the dome. Site AH01 is a ca. 5 by 5 m sized outcrop that marks the topographically highest point. This outcrop was sampled at four sub-sites (AH01.1 -AH01.4) from various orientations. Site AH02 was sampled on a ca. 10 m long road outcrop, located approximately 100 m northeast of site AH01.

Pokuhu lava flow (PH)
Pokuhu flow is the only lava sampled in this study that was sourced from the Tarawera Volcanic Complex. It is associated to the 14,009 ±155 cal yrs BP Waiohau eruptive episode and is most prominently exposed on the ~ 30 m high Tarawera Falls (

Magnetic mineralogy Methods
Rock magnetic experiments were carried out using a Variable Field Translation Balance (VFTB) at the University of Liverpool. Isothermal Remanent Magnetisation (IRM) acquisition, backfield coercivity of remanence, hysteresis and saturation magnetisation vs. temperature (Ms vs. T) measurements were made on a ca. 150 mg sample of each unit. The maximum field applied during in-field measurements was 800mT and thermomagnetic curves were measured to a peak temperature of 700°C.
The VFTB data were analysed using RockMag Analyser 1.0 (Leonhardt, 2006). In addition to displaying the data, this software calculates the hysteresis parameters including the magnetic coercivity (Bc), the saturation remanence (Ms) and the remanent saturation magnetisation (Mrs) from the hysteresis data after removal of a linear trend fitted to the closed parts of each loop, to correct for the paramagnetic component. The magnetic coercivity of remanence (Bcr) is calculated from the backfield IRM curves. Curie temperatures were estimated to be the temperature at which the second derivative d 2 Ms/dT 2 is a maximum (Tauxe, 1998).
The thermomagnetic curves allow assessment of the thermal stability of the magnetic remanence carriers. If identified, the onset temperature of alteration is of major importance for the analysis of thermal demagnetisation (THD) and palaeointensity data. The onset temperature can be best identified by measuring magnetic susceptibility in heating and cooling cycles to stepwise higher temperatues (e.g. see chapter 2). These experiments were not carried out on the rhyolitic samples addressed in this chapter. However the magnetic susceptibility (χ) was monitored alongside the THD experiments described in section 5.7. Variations in χ give important clues about potential mineral alteration and we thus discuss the results alongside the rockmagnetic data.

Results
The rhyolitic domes or flows sampled generally comprise pumiceous to crystalline lavas. At three sites (HN02, TC01, TC02) we sampled highly fractured volcanic obsidians. The texture can change from pumiceous to obsidian over a few meters, as is typical for rhyolitic lavas (Sigurdsson, 2000). For instance, site HN01 is on the outer rim of Hainini dome and comprises semi-crystalline pumice, while site HN02, located only a few hundred meters away comprised obsidian. These textural differences however do not affect the rock magnetic behaviour, suggesting that magnetic minerals formed prior to and remained in their present state after their emplacement.
The IRM acquisition curves of all samples studied saturate at fields lower than 250mT, indicative of a relatively "soft" magnetic remanence carrier such as titanomagnetites or (titano)maghemites ( Figure 5-2). The hysteresis loops indicate dominance of a ferro-/ferrimagnetic component and mostly yield hysteresis parameters traditionally associated with pseudo-single domain carriers (Dunlop, 2002). Exceptions are samples from Haroharo (HR) and Waiti (WT) lavas, which yield higher Bcr/Bc ratios but similar Mrs/Ms ratios to all other flows. The samples that fall into the PSD size range follow a theoretical trend for mixtures between SD and MD grains suggested by Dunlop (2002), which was also observed on the andesitic lavas described in chapter 3 ( Figure 5-3a). The actual magnetic grainsize can not be determined without additional experimental data, such as First Order Reversal Curves (FORC) (Roberts, 2000) or petrographic images.
In relation to the magnetic domain state, the magnetic coercivities vary between individual flows: The hysteresis loops measured on samples from Haroharo (HR) and Waiti (WT) lavas are extremely narrow and have coercivities between 3 and 8 mT.     (Tauxe, 1998). Type is the thermomagnetic curve type, defined in Figure 5- 1 Note: Site is the Site ID, ID the specimen ID, Tc1(HC), TC2(HC) are the Curie temperatures extracted from the heating curves, TC1(CC) and TC2(CC) are the ones extracted from the cooling curves. Type is the thermomagnetic curve type, defined in Figure  5

Demagnetisation methods and site averaging
On the assumption that the characteristic remanent magnetisation (ChRM) is a thermoremanent magnetisation (TRM), demagnetisation experiments were focussed on thermal methods. At least five specimens, from different samples per site were subjected to stepwise demagnetisation using a Magnetic Measurements Ltd thermal demagnetiser. Temperature intervals were selected on the basis of the thermomagnetic properties and samples were held for 30 minutes at each temperature prior to cooling to ensure equilibration. Progressive demagnetisation was conducted until less than 10% of the natural remanent magnetisation (NRM) remained. The low-field magnetic susceptibility of individual specimens was monitored after each heating step using a Bartington MS2 meter with the aim of identifying any magnetomineralogical alteration. We also subjected two specimens per site to alternating field (AF) demagnetisation using a Molspin AF demagnetiser. Starting from a treatment step with a peak field of 5 mT, the peak field was increased in 5 mT increments from 0 up to 30 mT, subsequently the step size was increased to 10 mT. AF demagnetisation experiments were carried out to a maximum peak field of 95 mT, by which point a characteristic remanence (ChRM) could be clearly identified from all specimens. Following thermal or alternating field treatment steps, all specimens were measured using an AGICO JR6A spinner magnetometer. All data analysis was carried out using Remasoft 3.0 (Chadima and Hrouda, 2006).
The remanence directions and intensities are displayed in vector component diagrams (Zijderveld, 1967). Statistical analysis was carried out following (Fisher, 1953). Flow mean and site means were calculated following the criteria and steps outlined below, and included the zero-field remanence measurements from palaeointensity experiments, where available: • All core orientations were referenced to geographic coordinates using the orientations measured using a suncompass in the field. The sun-compass orientations usually differed by up to 3° from the magnetic compass bearings and were accepted as more reliable. A table listing the individual sample orientations and differences between the magnetic and suncompass orientations can be accessed in the electronic appendices.
• With the exception Waiti (WT) lava flow, ChRM directions were calculated from the most stable (high Tb) component using principal component analysis (PCA) (Kirschvink, 1980), using a minimum of five data points. On flows where block rotation was identified as source for thermal overprinting, the ChRM directions were selected from the most recent (low Tb) component.
ChRM directions that did not lie within the cone of directions containing 95% of all directions (θ95) from a site were classified as outliers and rejected. For details on the interpretation of the demagnetisation results from Waiti (WT) flow specimens please refer to section 5.7.4.
• If more than one specimen was analysed from the same core, the ChRM directions extracted from each specimen were averaged into a sample mean.
• Site mean directions and associated cones of 95% confidence (α95) were calculated from all accepted sample means following Fisher (1953).
• If the site means of the same flow/unit agree at the 95% level of confidence or if the 95% confidence cones overlap it is assumed that all samples belong to the same population and an average for the respective unit is calculated from the individual sample directions. If a similar number of samples was available from each site, flow means were calculated from the sample means, otherwise flow means were calculated from the site mean directions.

Palaeomagnetic results
As discussed in section 5.5 we noted little variation in the rock magnetic properties between individual units and sites sampled, and the subsequent behaviour during demagnetisation experiments is similar for many of the sites sampled. The sampling details and palaeomagnetic results are outlined for each unit and the results summarised in Table 5

Haroharo dome (HR)
With values around 0.3 A/m the specimen NRM intensities fall well into the range of intensities measured on most other rhyolitic samples discussed in the following sections. During THD specimens from both sites (HR01 and HR02) lose their remanence in two major blocking temperature intervals (200-300°C and 500-600°C ), the two components being slightly in different direction ( Figure 5-6c). Separation of the two components is more difficult from the data of AF demagnetisation experiments, which are characterised by a rapid decrease in the remanence at peak fields lower than 20 mT, in agreement with the low Bc measured during hysteresis experiments ( Figure 5-6d).
There are two possible interpretations for the presence of a low temperature component: It could be either caused by a chemical remanent magnetisation (CRM), carried for example by a maghemite that inverts to a more stable endmember (e.g. hematite) between 200 and 300°C and thus causing a sudden decrease in the NRM intensity. However such an inversion or similar alteration effects would manifest in a sudden change of the magnetic susceptibility monitored alongside the heating experiment, which we did not find (see electronic appendix). We did not identify any secondary compositional phases in the thermomagnetic curves (section 5.5) and thus suggest that the sampled outcrop moved during cooling. In this case, the low

Hainini dome (HN)
Specimens from site HN01 demagnetise in two steps (200-400°C and 500-600°C), with the low temperature component showing a more easterly direction than the high temperature component. In contrast, all specimens from site HN02 show linear decay to the origin in the range of 400°C to 600°C (Figure 5-6f). The mean calculated from the low blocking temperature component from site HN01 is consistent with the direction obtained from the entire blocking temperature interval from site HN02 specimens. We therefore suggest that site HN01 may have been subject to some form of block movement during cooling, for example during endogeneous dome growth and calculate a best mean from the most recently magnetised components discussed above: Dec = 326.5°, Inc = -61.3°, α95 = 3.6, N = 16.

Waiti flow (WT)
All WT specimens demagnetise over a broad blocking temperature range between 200 and 580°C but demagnetise rapidly during AF treatment, with median destructive fields around 7mT. Specimens from both sub-sites carry overprints of random direction and the ChRM usually only comprises 10-20% of the total NRM ( Figure   5-7 a,b,c). The ChRM can be resolved more successfully from AF data than from the Data from each other specimen was fitted with the best fitting great circle (normalized) ( Figure 5-7d). A mean direction for the flow was calculated from the four ChRM directions and the best fitting direction on each remagnetisation great circle, determined using the iterative procedure outlined in by McFadden and McElhinny (1988). Best mean: Dec = 352°, Inc = -50.8°, α95 = 5.0°, N =8.
Figure 5-7: Representative Zijderveld, stereoplot and M/M0 plots from rhyolitic lavas for Waiti lava flow (WT). a) THD data for a site WT01.1 specimen b) THD data for a site WT01.2 specimen, c) AF demagnetisation data for a site WT01.2. specimen and d) stereographic projection of the endpoint ChRM selected from suitable specimens and remagnetisation great circles for specimen directions with strong overprints.

Rotoma flow (ROT)
The behaviour of samples from the individual sites and sub-sites differs. In general, all specimens lose their remanence primarily between 400 and 560°C, hower single component remanences were only identified on specimens from sub-site 1.1 while all other specimens are strongly overprinted (Figure 5-6g). The ChRM directions extracted agree at a sub-site level, however the individual site means disagree with each other and are thus treated separately ( Figure 5-9b). None of the mean directions agree with Tanaka et al.'s (2009) best estimate (NK19: Dec = 357.6°, Inc = -61.5°, α95 = 4.3°, N = 7). We suggest that the outcrops sampled were subject to some form of block movement along the flow-top surface or margins during cooling. A reliable direction estimate could therefore not be made for this flow.

Acacia Heights dome (AH)
All specimens demagnetise in a narrow blocking temperature interval between 450 and 550°C, and yield straight line decay to the origin with only minor viscous overprints ( Figure 5-6h) which is consistent with the thermomagnetic curves measured (section 5.5). In order to detect any potential post-cooling block movement on the dome top we calculated mean directions for each individual sub-site. Overlap between the 95% confidence cones is seen for all sub-sites other than AH01.4. Specimens from AH01.4 show an anomalously low inclination ( Figure 5-9d). While the source could not be identified based on field relations, we rejected all data from this sub-site, in consideration of the inconsistency between this result and those of other sub-sites close-by. The site means from AH02 and AH03 are steeper than the mean direction found at site AH01 and we suggest that some form of minor block movement occured after cooling. Assuming that the small-scale block movements noted were random, a site mean was calculated from the mean directions of individual sub-sites and our best estimate is Dec = 6.1°, Inc = -69.7°, α95 = 8.2, N = 6 (mean direction if calculated from all sample directions: Dec = 4.7°, Inc = -69.1°, α95 = 2.8, N = 31).

Pokuhu flow (PH)
With an average NRM intensity of ca. 5 A/m, samples from this flow are one order of magnitude more strongly magnetised than most other rhyolitic samples discussed in the sections above. In addition, all specimens from this flow showed unusual behaviour during thermal demagnetisation. While the NRM direction measured is of normal polarity, stepwise THD revealed a distinct reversed polarity component in the Tb range between 300 and 500°C, which causes a net increase in the magnetic remanence by approximately 20% (Figure 5-8a). Further demagnetisation results in a rapid loss in remanence in a narrow blocking temperature interval between 500 and 560°C. During     Tanaka et al. (2009) (NK01) 2. 8 -53.4 5.3 7 Flow mean calculated from the sample ChRM directions from one site.

Mean direction published in
Note: he tephra ages from Lowe et al (2013) are given in calendar years before present with 2σ confidence limits. Lat and Long are the site latitude and longitude, respectively. Dec is the ChRM declination and Inc the ChRM inclination. N/n/n0 = total number of samples included in the average/total number of specimens included/ total number of specimens demagnetised.

Experimental procedures
Palaeointensity experiments were carried out on samples from sites that showed consistent ChRM directions during THD and AFD, including Tapahoro, Hainini, Pokuhu and Acacia Heights lavas.
All palaeointensity experiments were carried out using the thermal palaeointensity method (Thellier and Thellier, 1959), following the Coe-type protocol (Coe, 1967) and including additional tail-checks (Riisager and Riisager, 2001). Some experiments on samples from Acacia Heights dome were also conducted using the IZZI type protocol (Yu et al., 2004). Please refer to chapters 2 and 4 description of the individual procedures. In this study we used a laboratory field of 50 μT during in-field steps, which is close to the average palaeointensity expected for the time-period. Each experiment was accompanied by susceptibility checks on individual specimens following each heating interval. The laboratory field was applied along the z-axis for each specimen, however the field was alternated between being applied parallel or antiparallel to the remanence direction, with the aim of enhancing potential pTRM tails (chapter 4). Temperature intervals were selected on the basis of behaviour observed during thermal demagnetisation experiments and, while we used large step sizes of 100 or 200°C up to a temperature of 500°C, we subsequently decreased the temperature increments to 15°C.

Data analysis and selection criteria
The data analysis was carried out using ThellierTool 4.11 (Leonhardt et al., 2004a) and the results were displayed in Arai plots, in which the NRM remaining after each zero-field demagnetisation step is plotted against the TRM gained during the corresponding in-field step. Palaeointensities were calculated from the best fitting slopes to selected data points. Based on the quality of the best fitting slope in the Arai plot, pTRM and tail checks we classify the individual specimen results into those that pass the more stringent TTA* selection criteria or the less stringent TTB* of Leonhardt et al. (2004a) and modified by Paterson et al. (2014). The individual statistical parameters were detailed in chapter 4 and calculated following the "Standard Palaeointensity Definitions v1.0" by Paterson et al. (2014) and the values are summarized in Table 5-3.

Palaeointensity results
Successful palaeointensity estimates were obtained on 12 out of 25 specimens, and include samples from Tapahoro (TC)

Tapahoro flow (TC)
All specimens yield a single component direction in the zero-field remanence steps and excellent repeatability in the ability to grow a TRM throughout the Tb spectrum as seen from pTRM checks. However the relation between NRM and TRM displayed in the Arai plots yields a slightly concave upwards shape. A linear fit to the low temperature segment (T < 500°C) of the plot yields a steeper slope than an approximation to data from the high temperature segment (T = 500-580°C) ( Figure   5-10a). The resultant palaeointensity estimates for all specimens average to 61.3 ± 2.8 μT for the low temperature, 31.1 ± 1.8 μT for the high temperature component and 45.5 ± 1.5 μT if data from throughout the entire Tb spectrum is included ( Figure 5-10b, Table 5-3). In all three cases the data pass the selection criteria set TTB*. The lowest palaeointensity estimate is consistent with the estimate of 31.0 ± 3.5 μT published in Tanaka et al. (2009), which like ours is based on the high temperature segment only.
However, we did not see indication of a separate low Tb directional component to justify the exclusion of any data. We also noted small differences between the remanence measured during zero-field and later tail checks and therefore suggest that this behaviour may be indicative of slight MD behaviour. Levi (1977) shows that for MD magnetites that do not alter during an experiment, an accurate palaeointensity estimate could be made from the two endpoint ratios within the Arai plots which correlates with the average calculated from the entire Tb spectrum. We therefore suggest that the intensity result provided by Tanaka et al. (2009)

Hainini dome (HN)
The palaeointensity results on most specimens pass at least the selection criteria TTB*, however in consideration of the block rotation discussed in section 5.7.3, all data from site HN01 were discarded. All successful specimens show a drop in the magnetic susceptibility at heating steps to temperatures higher than 500°C which correlates with a change in the relation between NRM and pTRM displayed in the Arai diagrams. We therefore reject all high temperature data from the calculation of intensities. Our final palaeointensity estimate of 51.7 ± 3.2 μT (N = 3/10) is placed near Tanaka et al.'s (2009) estimate of 58.1 ± 2.9 μT.

Acacia Heights dome (AH)
High quality palaeointensity data was obtained from 5 out of 6 specimens studied from Acacia Heights dome (AH). Experiments on this unit were carried out using the IZZI protocol. As discussed in section 5.7.6 the magnetic remanence is primarily carried within a narrow blocking temperature interval between 480 and 520°C. Samples do not alter, resulting in excellent repeatability of the ability to grow a pTRM throughout the experiments. No or insignificant scatter about the best fitting slope in the Arai diagrams was noted during IZZI experiments and the remanence direction shows single component decay to the origin ( Figure 5-10e). Our best mean intensity from five successful specimens is 63.3 ± 1.2 μT (N = 4/5).

Pokuhu flow (PH)
All palaeointensity experiments carried out on samples from Pokuhu flow fail. As discussed in section 5.7.7, all specimens from this flow carry a component with a remanence direction that is antipodal to the overall remanence direction in the blocking temperature interval 300-500°C, resulting in an increase of the NRM intensity in zerofield remanence and Arai plots by up to 20%. pTRM and tail checks fail at temperatures T > 500°C ( Figure 5-10e).
Figure 5-10: Representative Arai plots and corresponding vector component plots from the TVZ (insets) for specimens from the four volcanic units studied. a,b) Results from Tapahoro lava flow. Both Arai plots have a concave upwards shape and the intensity is therefore dependent on the fraction of the plot selected. c, d) results for Hainini lava dome. The Arai plots on both specimens yield a lower gradient slope for the high temperature component. However, based on a decrease in the magnetic susceptibility at temperature steps T > 500°C, this has been associated with thermal alteration. e) successful and f) rejected palaeointensity result for Acacia Height (AH) dome. The result was rejected due to the pTRM check failure at temperatures above 350°C g,h) Arai plots for Pokuhu flow show an increase in the NRM intensity in the blocking temperature range between 300 and 500°C, which corresponds to an antipodal directional component of magnetisation. Published palaeointensity result (Tanaka et al., 1994) 96.9 12.7 7 Note: Spec. ID corresponds to the specimen ID, Interv. is the temperature interval selected, n,f, β, q, MAD dCK, dpal, dTR, dt* are statistical parameters that were calculated following the "Standardized Palaeointensity Definitions" (Paterson et al., 2014), whereas MAD corresponds to the maximum angular deviation when anchored to the origin. Sel. criteria the set of selection criteria passed, as discussed in the text. Calculation of the best estimates: N/n number of successful experiments/ total number of experiments carried out. The mean directions and 95% confidence limits (α95) for all sites addressed in this study are displayed in a stereographic projection in Figure 5-9 together with our best estimate for each unit, and the previously published directions. Figure 5-11 displays the new data as a time series, in comparison to the previously published results, superimposed on the continuous record from Lake Mavora (Turner et al., 2015b), all relocated to 39.2°S, 175.54°E. Within the limitations mentioned previously and further discussed in chapter 6, the Mavora curve provides a reference frame for the directional and intensity values to be expected.
The α95's of the new flow mean directions range from 3 to 8° and the data dispersion is smaller than that of previously published data from the TVZ (Tanaka et al., 1994;2009) and within the range of the PSV study described in chapter 3. Although all  and Acacia Heights (AH) lavas fall significantly closer to the expected PSV values than was suggested by the original data (Tanaka et al., 1994;2009) (Figure 5-11).
The coordinates for site NK02 at Tapahoro flow listed by Tanaka et al. (2009) suggest a location that is approximately 200 m east of the margins of the flows and the sampled material may thus not have been in-situ (e.g. Table 5-2). We could not identify sampling location NK06 of Tanaka et al. (1994) at Acacia Heights dome but note that the original study was conducted with the primary aim of providing palaeointensity data and smaller emphasis may have been placed on the selection of in-situ sampling sites and interpretation of the directional data. Specimens from Waiti flow (WT) were affected by strong viscous overprints, described first by Tanaka et al. (2009) and also identified during this resampling campaign. These introduced some ambiguity in the data interpretation. We place confidence in our palaeomagnetic result which was calculated using the combined analysis of ChRM directions from four specimens and remagnetisation circles from four further specimens as described by McFadden and McElhinny (1988). For all three flows we suggest replacing the original data with our new directions.
Successful palaeointensity results were obtained from three lavas, which correspond to the units that Tanaka et al. (1994;2009) published results on. The overall success rate is 50%, comparable to that described on andesitic lavas in chapter 4 and the data dispersion at each site is 9% or less of its mean. High quality and consistent (σ = 1.2 μT) palaeointensity results were obtained for Acacia Heights dome (AH). The result falls into the range of palaeointensity values expected on the basis of the Lake Mavora record and other absolute palaeointensities obtained in this study, and is considered more probable than the very high estimate published previously (Tanaka et al., 1994). In contrast, the palaeointensity estimate made for Hainini (HN) dome agrees with the originally published data (Tanaka et al., 2009). The behaviour identified in the palaeointensity results on samples from Tapahoro (TC) flow also compares well to the data displayed by Tanaka et al. (2009).
Samples display a slightly concave shape in the Arai plots, as discussed in section 5.9.1, and we suggest selection of a different temperature range than proposed by Tanaka et al. (2009). The resulting palaeointensity estimate is significantly lower than the original value of Tanaka et al. (2009). While this estimate is adopted for the discussion below it should be regarded as tentative and used with caution during modelling or for comparative studies.
Figure 5-11: Palaeosecular variation described by the rhyolitic lavas between 5 and 15 kyrs BP. Our new refined data is displayed in red and compared to the previously published data (blue). Directional and intensity data that was previously of an anomalous direction is marked with an arrow pointing towards the refined values. The data is displayed in comparison to a continuous PSV record from Lake Mavora (black line with grey shaded error bar) (Turner et al., 2015b), which was relocated to the central North Island and provides a reference frame for the directions and intensities to be expected. Prior to its relocation the relative palaeointensity curve was scaled with a constant factor of 64 μT as suggested by Turner et al. (2015b).

Conclusions
This chapter presents the results of a palaeomagnetic study on seven Holocene rhyolitic lavas from the Okataina and Taupo Volcanic Centres. Data for most units were previously published in two studies by Tanaka et al. (1994;2009).
Recently available tephrochronology studies have enabled us to update the age control of each unit sampled, leading to revisions of up to 1000 years. The results suggest new palaeomagnetic directions on four flows, two of which gave unexpectedly easterly or westerly directions in the original publications. In contrast, an atypical westerly direction on a ca. 7.9 ka lava from the OVZ was confirmed as accurate. Successful palaeointensity estimates on three flows fall into the range of palaeointensities measured on flows from the Tongariro Volcanic Centre (TgVC).
While one estimate agrees with the data previously published, we suggest adopting the new palaeointensity values for lavas dated at ca. 11 and 5.5 ka.

Introduction
The main goal of this study was to build a sequence of discrete palaeomagnetic secular variation records and to review previously published data for the Holocene time-period in New Zealand from volcanic rocks of the Taupo Volcanic Zone.
Chapters 3, 4 and 5 presented the results of a comprehensive palaeomagnetic study on the volcanoes of the Taupo Volcanic Zone. Chapter 3 presented sampling strategies and directional results on andesitic lavas from the Tongariro Volcanic Centre (TgVC).
Chapter 4 complemented this with a detailed rock magnetic and palaeointensity study on the same flows. Besides collecting new palaeomagnetic data, in chapter 3 the ages of six lava flows from the TgVC were refined by comparison of their palaeomagnetic directions with the most recently published continuous lake sediment record from New Zealand (Turner et al., 2015b). Chapter 5 focussed on a palaeomagnetic study of rhyolitic lavas, primarily from the Okataina Volcanic Centre (OVZ).
Palaeomagnetic data on most of the sampled units from the OVZ had previously been published (Tanaka et al., 1994;Tanaka et al., 2009) and this study reevaluated and revised the early palaeomagnetic datasets from the area.
In this chapter, all new and existing palaeomagnetic datasets from Holocene (<15 kyrs BP) volcanic materials are compiled and the value of the dataset for regional and global PSV studies is then discussed. Future avenues for on-going and new research are presented. Table 6-1 is a compilation of all discrete palaeomagnetic data from volcanic material younger than 15 ka from New Zealand for which reasonable age control is available. Key datasets are the revised directions and intensities on rhyolitic lavas from the Taupo Volcanic Zone, presented in chapter 5, the data obtained from andesitic lavas from the TgVC, presented in the chapters 3 and 4, and other published results described below.

Dataset
The OVC sampling campaign included a number of units for which palaeomagnetic data had previously been published by Tanaka et al. (1994;2009 Robertson (2007). We also include the palaeointensity results from an obsidian from Mayor Island in the Bay of Plenty described by Ferk et al. (2011) and dated at 8800 ± 200 cal. yrs BP (Buck et al., 1981recalibrated using SHCAL13, Hogg et al., 2013. In addition we also include the palaeomagnetic results of two earlier studies which were conducted with a different focus: Robertson (1986) studied the basaltic lavas of Rangitoto Island, Auckland. At the time these lavas were thought to represent an extended period of activity, but more recent work (Needham et al., 2010) suggests just two phases of eruption at 504 ± 5 cal yrs BP and 553 ± 7 cal yrs BP. We assign an age range of 529 ± 31 cal yrs BP, which encompasses both eruptive phases and the associated age uncertainty and use the average palaeodirection of Robertson (1986) (Dec = 0.2°, Inc = -61.7°, α95 = 1.8°, 92 specimens from N = 11 sites). McClelland et al. (2004) obtained a comprehensive palaeomagnetic dataset from 44 sites sampled from the Taupo Ignimbrite (1720 ± 10 cal yrs BP after Lowe et al. (2013)).
Although the primary aim of this study was to gain insight into the emplacement temperatures, the ChRM directions extracted also provide an excellent record of the palaeomagnetic field. The dataset provides a mean direction for each site sampled, using the ChRM directions extracted from the low Tb range of a maximum of 13 specimens per site. The within site ChRM directions scatter widely and the site α95's average 13°, however the site mean directions are relatively well clustered.
Not included in this compilation are the results of a number of earlier palaeomagnetic studies. In a statistical study of secular variation in New Zealand, Cox (1969) sampled several units of Holocene age, which returned predominantly easterly declinations and inclinations close to the GAD value. Cox (1969) however made little use of demagnetisation techniques, and so the similarity to the present day field may reflect overprinting. Downey et al. (1994) reported palaeomagnetic directions from several series of flows on Egmont volcano (Mt Taranaki). They obtained predominantly easterly declinations, and inclinations between -50° and -70°, but, with little stratigraphic age control, it is presently difficult to incorporate these results into our compilation.

Addressing ages and age uncertainties
In order to be able to compare the discrete data with each other and with continuous records, precise and comparable age controls are required. The age information provided in Table 6-1 is based on three different radiometric and correlation methods: The andesitic lavas from the TgVC were dated either by 40 Ar/ 39 Ar dating or from under-or overlying, radiocarbon-dated tephra layers (Conway et al., 2016;Hobden et al., 1996;Lowe et al., 2013;Topping, 1974). Within the Holocene timeperiod the 40 Ar/ 39 Ar method has limited precision due to the slow decay of 40 K, and overall low potassium content in the lavas. Stratigraphic controls, on the other hand, are limited by the time intervals between the emplacements of over and/or underlying tephra beds. All independent age controls on these lavas thus carry uncertainties in the range of 2 -3000 years. In chapter 3 we refined these age brackets by amalgamating field relations and a comparison of our new palaeomagnetic directions with the continuous PSV record from Lake Mavora (Turner et al., 2015b), for which age control is available from the Bayesian modelling of sedimentation rate between 28 independent radiocarbon age estimates on leaf fragments throughout the record. The independence of such continuous records and discrete records that have been dated by comparison with them in this way is questionable, for example if both datasets are included in the same modelling exercise. For future reference therefore, we include in Table 6-1 both the palaeomagnetically-refined, more precise ages and the independent but less precise sets of age control..
Confidence in the palaeomagnetically-refined ages is further increased both by supporting field observations and agreement of our discrete palaeointensity data with the scaled relative palaeointensity record from Lake Mavora. In the following discussion, we therefore adopt the palaeomagnetically refined ages for the TgVC units.
Age controls on the rhyolitic lavas or other deposits included in our compilation were usually provided by correlation with distal tephra deposits. Most of these tephras are dispersed widely over the North Island in New Zealand and form distinctive marker beds in sedimentary environments. Once continuous PSV records have become available from this region, a direct correlation with the discrete palaeomagnetic data along the tephra horizons may be possible. To this end we adopted the radiocarbon ages from the most recent revision of the tephra ages and stratigraphies in New Zealand by Lowe et al. (2013). Lowe et al. (2013) calculated each tephra age from multiple 14 C ages of organic matter that was over and/or underlying the respective tephra horizon.
The standard deviation thus encompasses an unknown time period prior and after emplacement of the tephra. In contrast to 40 Ar/ 39 Ar, 14 C ages are affected by fluctuations in the atmospheric 14 C concentration, which affects the measured ratio between radioactive isotope 14 C and the stable isotope 12 C. Lowe et al. (2013) referenced the ages cited in this study to calendar ages, measured prior to 1950 AD, by correlation with the calibration curve SHCAL04 (McCormack et al., 2004). For the Holocene time period there is little difference between SHCAL04 and the more recent SHCAL13  which was used in calibrating the age model of the Lake Mavora sediments (Turner et al., 2015b). Calibration is necessary in order to compare the 14 C-based age estimates of the rhyolitic flows with absolute 40 Ar/ 39 Ar ages, palaeomagnetic ages and the Mavora timescale itself.
Calibration to calendar years BP can result in a shift of up to 900 years from the uncalibrated 14 C ages listed in the initial publications (Tanaka et al., 1994;2009).
Whereas the uncertainty quoted in an uncalibrated 14 C age arises primarily from experimental counting uncertainties, calibration introduces additional uncertainty associated with the calibration curve itself. Occasionally ambiguities arise, where a given 14 C/ 12 C ratio occurred more than once during the Holocene, and two (or more) ranges of calibrated ages are possible. In general however, the tephra ages carry much lower uncertainties than the 40 Ar/ 39 Ar ages, in the range of 50-250 years.
It should be noted that the assignation of ages assumes a direct correlation of the proximal deposits, including the lavas and volcaniclastic materials sampled for palaeomagnetic study with the distal tephra in the sequences studied by Lowe et al. (2013). An additional source of age uncertainty of individual proximal units arises from the duration of individual eruptive sequences. Nairn (2002) found that 14 C ages of charcoal from proximal deposits of the OVC usually fell within 100 to 300 years of the distal tephra ages, which provides an estimate to the additional uncertainty to be expected due to this. Okataina Volcanic Centre, Tanaka et al. (2009) and Robertson (2007) 6.5 Integration with regional datasets and global field model 'pfm9k' Comparison to the Lake Mavora PSV record (11,250 yrs BPpresent) Figure 6-2 displays all discrete palaeomagnetic data from New Zealand listed in Table   6-1 superimposed on the recently published continuous lake sediment record from Lake Mavora, in the South Island of New Zealand, covering the last 11,500 years BP (Turner et al., 2015b).
In order to accommodate the differences in the palaeomagnetic record expected due to the geographic separation between the sampling sites, all data were migrated to a common location within the central North Island (39.2°S, 175.54°E) using a VGP transformation (Noel and Batt, 1990). This transformation is strictly valid only in a dipolar field. However experiments on the relocation of field directions calculated from the gufm1 model (Jackson et al., 2000), which is based on historical data spanning the last 400 years, suggested that the migration error in this region and over a maximum distance of 1000 km is not expected to exceed 1.5° (pers. comm. Turner) which falls into the existing uncertainty of our data.
The discrete records summarized in Table 6-1 range between extremes of 326.5° west to 20.8° east and inclination extremes of -46.3° and -81.4°. They average to Dec = 2.9°, Inc = -62.4°, θ-63 = 9.9°, α-95 = 3.6°, N = 23 (Figure 6-1) (or including only data < 11.5 kyrs BP: Dec = 2.4°, Inc = -62.1°, θ-63= 9.2°, α-95= 4.2°, N =14).This mean is slightly but insignificantly east and steeper than the direction of a geocentric axial dipolar (GAD) field at the site (Dec = 0°, Inc = -58.5°). It is commonly believed that in a temporarily well-distributed, unbiassed dataset, averaged over a sufficient time-interval (probably 10,000 -100,000 years), secular variation averages to a geocentric axial dipolar field and the GAD direction would be an appropriate assumption (Merill et al., 1998;Turner et al., 2015b). However in chapter 3 we suggested that many of the Holocene flows from the TgVC were probably emplaced during a relatively short time-period during which the declination was significantly east of north. For instance the mean direction calculated from the discrete data listed in Table 6 Table 6-1 still lacks discrete PSV data in the time-period 2-5 kyrs BP. Figure 6-1: Distribution of the Mavora data (blue) and all discrete palaeomagnetic directions presented in Table 6-1 (red) and associated angular standard deviation θ63. All data were relocated to 175.5°E, 39.2°S. Projection: Wulff (equal angle).
Volcanic and sedimentary records complement each other in that, while the volcanic data provide absolute and instantaneous records of the palaeomagnetic field, the sequences of sedimentary records are quasi-continuous. The sedimentary records however are unavoidably smoothed to some extent and often display erroneously shallow inclinations, as result of the depositional process and later sampling procedure.
Furthermore these records only provide relative palaeointensity and declination as they are not oriented azimuthally and thus require calibration with independent and absolute records. The Lake Mavora directional curve was oriented based on a correlation with the gufm1 global field model (Jackson et al., 2000;Turner et al., 2015b). Overall, the discrete data presented in Table 6-1 agree with these features extremely well. A direct and independent comparison between the directional record of the TgVC lavas and the Lake Mavora Curve is difficult because the independent age controls on the TgVC lavas are subject to large uncertainty and the refined ages are based on a comparison with the Lake Mavora Curve itself. However, as discussed in chapter 3, many of the pre-5 ka lavas from the Tongariro Volcanic Centre (TgVC) describe a strongly easterly declination that compares well to the field recorded by the Mavora lake sediment prior to 8 ka. Field evidence and its 40 Ar/ 39 Ar age suggest that Taranaki Falls flow (TF) is the oldest of a sequence of flows with easterly directions (chapter 3) and the flow has a much shallower inclination than the younger flows of Delta Corner (DC) and Bruce Road (RP) ( Figure 6-2) which agrees with the behaviour of the field at around 8-9000 yrs BP.
All other volcanic records resampled in this study or revised in Table 6-1 have independent and more precise ages that allow a more rigorous comparison. In the late Holocene, a remarkably good match is seen between the palaeomagnetic The palaeointensity datasets also correlate well. All discrete palaeointensity data fall well into the range of swings suggested by the Mavora Curve, after applying the calibration suggested by Turner et al. (2015b). As discussed in chapter 4, the discrete palaeointensity record from the Tongariro Volcanic Centre (TgVC) compares well with the relative palaeointensity curve, within the age brackets obtained by palaeomagnetic dating using the directional records in chapter 3. Further the data follow an overall trend from lower to higher palaeointensities throughout the Holocene, as is also described by the most recent global model pfm9k (Nilsson et al., 2014). The discrete data from the rhyolitic units sampled in this study or revised in Table 6-1 support the additional short periodicity swings described by the Lake Mavora Curve. The good data fit is remarkable considering the difficulties discussed in chapters 4 and 5 with the extraction of palaeointensity data from lavas and that the relative palaeointensities of the sediments and the absolute palaeointensity data were obtained during two significantly different physical methods.
Overall the agreement between the discrete directional and, where available, intensity records with the Mavora Curve suggest that both the discrete and the continuous records are reliable recorders of the palaeomagnetic field and support Turner et al.'s (2015b) observations on geomagnetic field behaviour. This outcome also supports the orientation and inclination corrections applied by Turner et al. (2015b) to the Mavora raw data and the proposed calibration of the relative palaeointensity curve.
This outcome also enhances our confidence in the palaeomagnetic ages of the TgVC flows, which were refined using the Mavora directions (chapter 3).  Table 6-1 are displayed in comparison to a continuous PSV record from Lake Mavora (Turner et al, 2015b) and the GAD direction, the palaeointensity data are displayed also in comparison to the global field predictions of model pfm9k.1a (Nilsson et al, 2014). All data was relocated and the model prediction calculated at 39.2°S, 175.54°E. The uncertainties for the Lake Mavora data (grey shaded envelope) are displayed, and the discrete datasets have been calculated following the Fisherian formulae for 95% confidence in the mean direction (∆Inc = α95, ∆Dec = α95/cos(Inc)), and correspond to the standard error in intensity.

Scaling of the amplitudes of the Lake Mavora PSV record
As discussed above, sedimentary records are almost always affected by some form of amplitude reduction (e.g. Barton and McElhinny, 1881;Turner and Thompson, 1982).
Amplitude reduction was also suggested for the Mavora record (chapter 3) based on a comparison with the discrete datasets from the TgVC. Correction of the amplitude reduction is required to understand the full scale of variations in Earth's magnetic field in this region and will also improve the accuracy of further palaeomagnetic dating studies. The derivation of a suitable scaling factor however is difficult. Earlier palaeomagnetic studies (e.g. Turner and Thompson, 1982) used a peak-to peak comparison of several sedimentary curves with a different data distribution from the same region to estimate a scaling factor. This scaling factor can be used to correct the amplitude reduction present within a sedimentary record. This approach however relies on the assumption that at least one of the continuous datasets included in the comparison describes the full scale of PSV swings.
An alternative approach may be to compare the sedimentary PSV datasets to the absolute directional estimates from volcanic materials. A scaling factor could then be calculated by either a) Direct matching between individual data points from volcanic materials with the Mavora Curve or by calculating a least squares fit between the continuous and discrete data. The uncertainty in the estimate of the scaling factor will be large when the independent ages of the volcanic data have excessively large uncertainties, such as the TgVC lavas in this study, and the directional data thus have a number of potential tie points within the Mavora Curve.
b) Matching the dispersion of the PSV described by the Mavora record to the statistical distribution of the volcanic dataset. This approach requires a detailed analysis of the statistical distributions of the sedimentary dataset to be scaled and the volcanic records. It also requires an investigation of experimental noise in both datasets. A preliminary statistical analysis is conducted below and difficulties of a scaling approach using the volcanic dataset of this study outlined.
Many of the volcanic records listed in Table 6-1 are not different from the Mavora Curve at the 95% confidence level (e.g. the 95% confidence limits overlaps the confidence limit of the Mavora Curve), which to a first order suggests validity of the data comparison conducted above and the results of palaeomagnetic dating conducted in chapter 3. However, the flow mean directions often suggest overall broader swings (Figure 6-2). In Figure 6-3 we display a) the distribution of the Mavora data about their mean, b) the distribution of all volcanic data records that sample the same time-period (< 11.5 kyrs BP) as the Mavora Curve about their mean c) the distribution of all volcanic data listed in Table 6-1 about their mean direction. The relatively large data volume of the Mavora Curve (N = 224) is roughly Fisher-distributed, and has a dispersion of 5.4° after its relocation to 39.2°S, 175.2°E (Mavora θ63 = 4.8° prior to its relocation). In contrast the discrete datasets displayed in Figure 6-3b and Figure 6-3c have dispersions of θ63 = 9.1° and θ63 = 9.9°, respectively. Inter-site directional scatter or other noise, resulting from local magnetic anomalies in the field, sampling and experimental procedures are likely to enhance the dispersion of the records from volcanic materials but cannot account for the large difference between the Mavora Curve and the discrete dataset.
In a preliminary approach a scaling factor for the amplitude of the Mavora record could therefore be derived by matching the statistical distribution of the Mavora record to the one described by the volcanic dataset, including data from materials < 11.5 kyrs, which implies a scaling factor of ca. 1.7. We note however that the relatively small number of volcanic data included in Figure   6-3b do not present a large enough sample of the PSV distribution to form a distribution and hence to make an accurate estimation of its statistical parameters, also seen from the difference of its mean from the mean of the Mavora record. Inclusion of additional data > 11.5 kyrs (Figure 6-3c) would widen the dataset but would be conducted under the assumption that PSV prior to 11.5 ka followed similar patterns as in the later Holocene. Furthermore, as discussed earlier in this chapter, the volcanic dataset primarily samples two periods of high amplitude secular variation and it may therefore not be representative for the entire record. Rigorous statistical analysis thus requires a much larger and temporarily well distributed dataset. .5 kyrs about their mean c) Distribution of all discrete PSV data (<15 kyrs BP) about their mean. The red lines represent the angular bandwidth about the mean that contains 63% of all data (θ63).

The Lake Pounui record (2400 yrs BP -250 yrs BP?)
Not discussed in this thesis so far was the earliest continuous PSV record from Lake Pounui near Wellington in the lower North Island (Turner and Lillis, 1994), a directional curve at the time of publication thought to cover the past 2500 yrs BP. The Lake Pounui record describes a broad swing towards westerly declinations and back to the current easterly declination and from relatively shallow to steep inclinations in the late Holocene, features also described by the Mavora Curve. However in comparison to the Lake Mavora Curve the Pounui record appears shifted further back in time ( Figure 6-5). Turner et al. (2015b) suggested that the age model of the Pounui record, which was based on four uncalibrated radiocarbon ages, is too old by several hundreds of years. This was probably caused by inclusion of older organic material, with an inbuilt 14 C/ 12 C age, in the sediment samples used for radiocarbon dating. In recent efforts to integrate the Mavora, Pounui and a to-date unpublished sedimentary record into a New Master Curve for New Zealand (Corkill, 2015;Turner et al., 2015a) a revised age model for the Pounui sedimentary record was put forward. In the following we demonstrate that an age model can also be provided by correlation with the volcanic records younger than 2500 yrs BP, for which radiometric age controls are available (section 6.4). Independence in the age models of the continuous records is a requirement for stacking exercises, for example during the creation of a Master Curve.
In a preliminary approach and, starting from the most recent, we match each palaeomagnetic direction from the volcanic dataset with a point on the Pounui Curve. At its youngest end the Pounui record is dated to 250 14 C yrs BP and the easterly directions from the Tarawera Basalt (NK19) and Central Crater (CC) flow are probably placed at or above its upper limit. The youngest matches were identified for the Rangitoto basalt (RT) and Kaharoa rhyolite (KH) which correlated best with the mean directions of the Pounui record at 825 and 1075 14 C yrs BP in its original age model, respectively.
Going backwards in time, the Pounui record describes a rapid swing towards westerly directions and shallow inclinations and the declinations peak at around 1500 and 1700 14C yrs BP (Pounui age model). From this point the declinations return to easterly directions at the lower end (~2400 14C yrs BP, Pounui age model) of the record. If this age model was accurate this would not allow for the inclusion of the easterly palaeomagnetic direction of the Taupo Ignimbrite (Dec = 3.9, Inc = -52.2°, α95 = 2.4°), which has a radiocarbon age of 1720 ± 10 cal. yrs BP ( Table 6-1, Figure   6-5).
We thus suggest that the direction of the Taupo  The age calibration suggests that at its lowest end the original age model is up to 900 years too old (Figure 6-4). Figure 6-4: Calibrated age model for the Pounui record based on a correlation of the palaeomagnetic record with discrete data points from two volcanic lavas for which age constraints are available (section 6.4) and inference that the Taupo eruption preceded the earliest record. The x-axis displays the original 14 C ages of the Pounui record, on the y axis we display the calibrated ages. Figure 6-5: Comparison between the Mavora Curve and the Pounui record prior to the correction of its age model, the Pounui record after its correction and the discrete data discussed in the text. All data was relocated to 39.2° S, 175.5° E. The revised Pounui Curve reproduces the features of the Mavora Curve between ca. 1700 cal. yrs BP and present.

Effects of the new data on global field model 'pfm9k.1a'
With the aim of testing the impact of the new discrete data on the most recent global field model pfm9k.1a, which incorporates sedimentary and absolute directional and intensity records (Nilsson et al., 2014), Andreas Nilsson kindly re-ran the modelling procedure including the discrete data presented in this chapter. Figure 6-6 compares between model pfm9k.1a a) as it was published in 2014 (green), including Turner and Lillis (1994) continuous record from Lake Pounui and the discrete data of previous studies by Tanaka et al. (1994;2009) and Robertson (1986)

b) after
replacing the Pounui record with the Mavora Curve (red), c) after replacing all previously included datasets from New Zealand with the Mavora Curve and the revised or new data presented in this thesis (blue). In this preliminary approach we used the palaeomagnetic ages from chapter 3 as age control for the TgVC lavas (Table 6-1). We use these refined ages to improve the usability of the data for the modelling approach, where age uncertainties are translated to palaeomagnetic uncertainties (e.g. Korte et al., 2005).
When the Mavora record is included in the calculation of pfm9k.1a the overall model reproduces its details very well, as would be expected as the only other data from the SW Pacific region are five lake sediment records from Australia (Barton and McElhinny, 1881;Constable, 1985;Constable and Mc Elhinny, 1985), more than 2500 km north-east of New Zealand. The most prominent changes to the original model are visible in the late Holocene (2000 yrs BP -1000 yrs BP), where the original model pfm9k.1a was primarily constrained by the Lake Pounui record (Turner and Lillis, 1994), which has an age model that is probably too old (section 6.5.3). A prominent swing towards easterly declination and shallow inclinations described by the Mavora Curve in the early parts of its record (~9000 yrs BP), has a significant effect on the model as well. However most of this high amplitude signal is earlier than the lower end of model pfm9k.1a. As discussed earlier, the discrete data suggest even higher amplitude swings than described by the Mavora Curve. It is therefore not surprising that inclusion of the additional discrete data to the modelling procedure results in an even stronger bias of the model towards easterly directions at its earliest end. However, overall we note that in comparison to the Mavora Curve the discrete data has only minor impact on the model, most probably due to the smaller data volume, its wide temporal and spatial distribution (pers. comm. Nilsson), the overall good agreement with the Mavora Curve and the also the fact that the ages of the TgVC lavas are not independent of the Mavora Curve. Single and extreme data points often result in short wavelength biases of the model towards the respective data. This outcome has a number of implications: • A larger dataset from the region is required to ensure that outliers are not overly weighted during the modelling procedure.
• To decrease the impact of single data points but to place higher emphasis on the overall scale of PSV described by the absolute records we suggest the implementation of some form a pre-scaling procedure for the amplitudes (e.g. section 6.5.1) for the continuous records.
A pre-scaling procedure may also allow for the inclusion of palaeomagnetic data from the TgVC lavas with their independent and large uncertainties. Removal of these datasets from the modelling procedure would be unfortunate as following the data analysis presented in chapters 3, 4 and 5 we place overall higher confidence in the palaeomagnetic data obtained from the andesitic lavas from the TgVC than in that of many of the rhyolitic domes or flows. This is because the TgVC lavas were usually clearly in-situ, showed coherence of palaeodirections over wide site-spreads and were rarely affected by VRM or CRM overprints. In contrast on some of the rhyolitic lava flows or domes coherence could only be established through wide spread sampling campaigns. Figure 6-6: Declination, inclination and intensity records for 39.2°S, 175.5°E calculated from the global field model pfm9k.1a in its original and published form (Nilsson et al., 2014), after inclusion of the Mavora Curve and after addition of the discrete data discussed in this chapter.

VADM reconstruction
Comparison of palaeointensity data from New Zealand with global datasets is possible by comparison of virtual axial dipole moments (VADMs) or virtual dipole moments (VDMs) with global averages. A VADM describes the moment of the geocentric axial dipole that would produce the observed palaeointensity at the latitude of the site. It is calculated using: where F (T) is the palaeointensity measured, R (m) Earth's mean radius, μ0 (m*kg*s -2 *A -2 ) the permeability of free space and θs the co-latitude of the site.
Temporal and global averages of VADM aim to average out variations in the nondipolar field (e.g. Yang et al., 2000) and thus present a record of the evolution of Earth's dipole moment. Deviations from the global average, observed in single studies may bear information about the effects of the non-axial and non-dipole components of the field in a particular region. Figure 6-7a displays the virtual axial dipole moments (VADM) calculated from the discrete palaeointensity records presented in Table   6-1 and those from archaeointensity data from the SW Pacific Islands (Stark et al., 2010) in comparison to a global VADM reconstruction by Knudsen et al. (2008).
The latter is based on discrete palaeointensity data from the GEOMAGIA50 database (Donadini et al., 2006). Knudsen et al.'s (2008) curve describes a long wavelength swing prior to 7,000 years BP, from when it increases towards a prominent high at around 3,000 years BP. From 1,000 yrs BP the VADM reconstruction describes a rapid decrease towards the present day axial dipole moment of 7.72*10 22 Am 2 (Thébault et al., 2015) With one exception (Acacia Heights dome) our pre-5 ka data compare well with the VADM reconstruction. Considerable differences are however seen between the global reconstruction and the discrete data from the Pacific region in the late Holocene. For instance, Stark et al.'s (2010) archaeomagnetic data (2500 -4000 years BP) generally displays significantly below the VADM average. In contrast, the palaeointensity results on the young flows from New Zealand, for example Central Crater flow (palaeomagnetic age 300 ± 200 yrs BP) and the (Tanaka et al., 2009) yield VADMs significantly above the global VADM reconstruction. The high palaeointensity recorded at Central Crater (CC) flow is accompanied by a steep inclination, which suggests an offset of the south magnetic pole towards New Zealand. From around 1700 AD global models show high magnetic flux at the core-mantle boundary (CMB) in the South Pacific (Nilsson et al., 2014) with tilting of the magnetic poles towards the pacific hemisphere. South magnetic pole movement towards lower latitude has continued during the last 100 yrs BP, as recorded in observational data (e.g. Finlay et al., 2010).
Figure 6-7: VADM's calculated from all intensity data from NZ volcanics and Pacific Island pottery (Stark et al., 2011) in comparison to the global VADM reconstruction of Knudsen et al. (2008).

Future Work
A spatial and temporal extension of the present dataset would enhance our understanding of the evolution of the palaeomagnetic field in New Zealand and the SW Pacific region. The discrete dataset presented in this thesis lacks data in the timeperiod between 5000 and 2500 yrs BP, due to the relative quiescence the Okataina Volcanic Centre and a dearth of well-dated materials from the Tongariro Volcanic Centre in that time-period. A priority of further studies on Holocene volcanic materials in New Zealand should therefore be an extension of the sampling campaign to other volcanic areas. For that purpose the inclusion of other materials than lavas, such as welded ignimbrites may be explored. Likewise, discrete and continuous datasets from the wider Pacific region are required in order to produce more robust field models and to understand the non-dipolar contributions to the field particularly in the late Holocene.
A major goal of future studies lies in the combination of the new datasets, including discrete data from volcanic and archaeomagnetic materials, and continuous records from New Zealand into regional field models (e.g. Alfheid, 2014;Ingham, 2009) and a PSV Master Curve for New Zealand (e.g. Corkill, 2015;Kinger, in prep;Turner et al., 2015b). A PSV Master Curve will be beneficial not only to an understanding of Earth's magnetic field but will provide important constraints for later palaeomagnetic dating studies. The potential of palaeomagnetic dating to refine age controls on lavas that are associated with large uncertainties was demonstrated on seven flows from Mts Ruapehu and Tongariro in chapter 3. Additional palaeomagnetic studies in particular in the northern part of the Tongariro Volcanic Centre, where the cone-building stratigraphic sequences are only loosely defined from under-and/or overlying tephra marker beds (Hobden, 1997) may be able provide additional age constraints and highlight individual effusive episodes.
This study presents the results of a comprehensive palaeomagnetic study on Holocene lavas from the Taupo Volcanic Zone, with a focus on the Tongariro and Okataina Volcanic Centres. The new dataset includes: • New palaeomagnetic directions from twelve units from the Tongariro Volcanic Centre, six flows from the Okataina Volcanic Centre and a direction from a rhyolitic dome near Taupo in the central North Island in New Zealand.
• Ten new palaeointensity estimates on flows from all three volcanic complexes are obtained using the thermal Thellier-type palaeointensity method and, for the andesitic lavas, also the microwave palaeointensity technique.
• The new data, and data previously published from the region are integrated into a new discrete PSV data compilation for New Zealand, including 24 high quality directional data and ten palaeointensity estimates. Each data carries independent high resolution 40 Ar/ 39 Ar age constraints from a parallel PhD study by Conway (2015)  Integration with continuous sedimentary records from New Zealand or the most recent global field model pfm9k.1a leads to the following conclusions/suggestions: • The new data compare well to the recently published sedimentary record from Lake Mavora (Fiordland, New Zealand) (Turner et al., 2015b) and support the features described therein. The overall data dispersion (θ63) of the data from volcanic materials is significantly larger than that of the Lake Mavora Curve, suggesting that the Mavora Curve is subject to amplitude smoothing. A comparison of the earliest secular variation record from Lake Pounui in the lower North Island, New Zealand (Turner and Lillis, 1994) with the Mavora Curve suggests that the age model of the earlier record is too old (Turner et al., 2015b). A tentative age model is presented in this study, based on a comparison with the independently dated volcanic dataset. The new model suggests that, at its lower end, the Pounui Curve is up to 900 years too old.
• The virtual axial dipole moments of the absolute palaeointensity data of this study and archaeomagnetic data from the south west Pacific Islands (Stark et al., 2010) are compared with a global VADM reconstruction by Knudsen et al. (2008). The datasets from New Zealand compare well to the VADM reconstruction in the early Holocene, but significant differences are visible in the late Holocene (ca. 3000 yrs BP until present). Above average VADM values recorded by our new data younger than 1000 yrs BP are in accordance with a tilting of the magnetic poles towards the Pacific region, visible in global field model pfm9k.1a (Nilsson et al., 2014) and from observational data (Thébault et al., 2015).
• Inclusion of all new datasets into the global field model pfm9k.1a, which was re-run by Andreas Nilsson, has a strong impact on the model, most probably due to the small volume of data previously available from the region. In comparison to the Mavora Curve the volcanic data has only minor effect, probably due to the smaller number of data available and the good agreement with the Mavora Curve. Suggestions as to how volcanic PSV records could be addressed differently during modelling procedures are made.
• A Bayesian based statistical Matlab tool (Pavón-Carrasco et al., 2011) is used to refine the independent radiometric age constraints on five lava flows from the Tongariro Volcanic Centre by comparison with the Mavora directional record. The new age estimates reduce the age uncertainties from between 2-3000 years to as little as 500 years.