Item

A dendroclimatic study of Libocedrus bidwillii hook. F. (Kaikawaka)

Xiong, Limin
Date
1995
Type
Thesis
Fields of Research
Abstract
This thesis demonstrates some of the potential of Libocedrus bidwillii Hook. f. (Kaikawaka) for dendroclimatological research by developing tree-ring chronologies and then using these chronologies to reconstruct palaeoclimates. In order to assist with the modelling of tree-growth and climate relationships (response function analysis), the annual nature of Libocedrus bidwillii growth was investigated. Results showed that seedlings of Libocedrus bidwillii were sensitive to temperature and soil moisture. Greatest growth was at high soil moisture and under a variable temperature regime. It was also found that there was an obvious seasonal variation in the growth of the seedlings. Such information allowed some confidence in the use of the species as a proxy-climatic indicator. Twenty-three tree-ring chronologies were developed from different areas of New Zealand. These included 12 new sites, 5 sites collected by other people but then updated and 6 sites that were not updated. Standardisation of the tree-ring series from each site used double detrending methods - ERH+SP67% (linear-Exponential or linear Regression or a Horizontal detrending plus SPline detrending fitted to 2/3 the length of the tree-ring series). This meant some long-term trends in the data were retained (i. e. greater than 120 year cycles) although this led to some reduction of the strength of the common signal in the chronology as measured by EPS (Expressed Population Signal) and SNR (Signal of Noise Ratio). The retention of long-term trends in the chronologies was thought to be important because some low frequency signals, which are longer than 120 years, are present in the climate data. Autocorrelation in the chronologies was removed by the ARSTAN program using the Aikaike Information Criterion (AIC) to determine the filter model. No significant autocorrelations were left in the residual chronologies produced by this method. Inter-comparison of the chronologies showed a highly consistent and significant pattern between most of the sites. There was little reduction in inter-chronology correlation with separation distance. However, there was a difference, or an effect, due to altitude. In general the response functions for the relationship between climate variables and ring-width in any given growing season showed a negative relationship between temperature for the prior growth months February, March and current December, while there was a positive response to temperature in September and February. There were three significant negative coefficients (previous March, April and August) and one positive (current February) for precipitation. The results of using principal component analysis (PCA) showed that all the 27 significant response function analyses could be divided into four groups. The response pattern in the four groups was similar for temperature but the rainfall response was more variable. The climate reconstructions were based on two groups of chronologies: eleven chronologies from all over New Zealand and a subset of only the three longest chronologies. Comparison of the climate data of different seasons with the two groups of chronologies was carried out using the "bootstrap" transfer function. The average February-March temperature and total March-April precipitation were finally selected as the reconstructed variables. Both of the groups reconstructed the hot years better than the cold years. The reconstructed temperature series were similar to all the earlier New Zealand dendroclimatic reconstructions. The warming and cooling periods, extremely warm and extremely cold years were identified and compared with some other sources of evidence and found to be highly consistent. This led to the conclusion that Libocedrus bidwillii is very useful as a high resolution palaeotemperature indicator. In the precipitation reconstruction, all the periodicities (both high and low) in the observed data were reconstructed. The dry and wet periods, severe drought and very wet years were identified in both precipitation reconstructions and also compared with other more limited sources of evidence.