The Climate Of New Zealand Reconstructed From Kauri Tree Rings: Enhancement Through The Use Of Novel Statistical Methodology

Abstract

Knowledge of past climates needs to increase to allow a coherent picture of current and past climate systems to be developed. Kauri tree rings are a significant resource that allows us to better understand past climates within New Zealand. The variation in the ring widths for kauri has previously been shown to be related to the El Ni˜no-Southern Oscillation (ENSO) phenomenon, where the evolving chronology variance can be interpreted as a reconstruction of the past activeness of ENSO events. Using this interpretation, the 201kauri master chronology (AGAUc10c), containing both archaeological and modern (living tree) source material of kauri ring widths sequences suggests that ENSO was less active in the 150s and activity appears to be increasing from the 150s to the present. Another key feature of the reconstruction is the inter-annual to decadal-scale periodicities. These interpretations rely on the assumption that the observed trends are a true representation of past climate variation. However, there are concerns that the change in the size and/or age composition of the cores making up the chronology may be affecting the observed trends. The primary aim of this research was to improve the quality of the AGAUc10c master chronology by investigating the influence of size and/or age on the climate response registered within kauri tree rings. This research asked two main questions: Is there a difference in the climate signal registered within near-pith material (rings formed when the tree was small (young)) when compared to mature growth material and can any such effect be mitigated through the removal of this near-pith material? Is there a difference in the climate signal registered within rings formed when the tree was very large or old compared to when the tree was mature? In addition, the influence of slow and fast growth phases on the chronology and the best chronology construction procedures with regard to minimising the influence of these growth phases were investigated. A concordance method was developed to detect the strength of evidence of differences between two subsets containing ragged arrays of dependent non-stationary time series. This method allows subsets of tree-ring data to be compared, such as a subset containing only near-pith material compared to a subset containing no near-pith material. Using the concordance method a difference was detected between the two subsets of data; only near-pith and no near-pith material. Furthermore, there were differences in the seasonal correlation functions between the chronologies produced from these subsets, suggesting that near-pith material has a significant weaker climate signal during the growing season than material further from the pith. The inclusion of the near-pith material therefore weakens the common climate signal and should be excluded from the dataset before constructing the chronology. There were only minor differences detected between chronologies comparing very old and mature material. Additionally, this very old material contains a similar strength of climate signal when compared to mature trees. Very old material should therefore be retained within the chronology. Tree ring series from kauri are identified as having the potential to contain localised periods of slow or fast growth. Series showing these features have the potential to influence the common climate signal registered within the chronology and should be removed from the dataset before the chronology is constructed. It was also recommended, to minimise the influence of these periods of slower or fast growth, that the chronology be constructed by: Fitting a growth trend to the raw ring widths using a spline. Constructing indices by dividing the raw ring widths by the fitted growth trend. Combining the indices using either the median or a robust mean. The final part of this study was to investigate the implications of making the changes recommended above on the ENSO interpretation, by comparing the evolving chronology variance (STD) from the chronology produced based on the above recommendations with the AGAUc10c master chronology. Overall the trend in the evolving STD remained substantially unchanged. The interpretation of the increasing activeness of the ENSO from the 150s to the present is still observable, although it is no as pronounced as previously inferred. The decade-scale periodicity are largely unaffected. The quality of the master chronology has been improved, as we can confidently say that the observed trends are not solely a consequence of changes in the size and or age composition of the chronology.