Wavelet analysis of flowering and climatic niche identification

Abstract

This chapter discusses wavelet analysis which is a robust statistical method capable of handling noisy and non-stationary data which phenological time series often are. We used a maximal overlap discrete wavelet transform (MODWT) analysis to examine the flowering records (1940-1970) of E. leucoxylon and Eucalyptus tricarpa, E. microcarpa and E. polyanthemos. We identified four subcomponents in each flowering series: characterised as a non-flowering phase, duration, annual and intensity cycles. A decreasing overall trend in flowering was identified by the MODWT smoothed series. Wavelet correlation found the same contemporaneous effects of climate on flow-ering for E. leucoxylon and Eucalyptus tricarpa, and for E. microcarpa and E. polyanthemos. Wavelet cross-correlation analysis identified the cyclical influence of temperature and rainfall on peak flowering intensity. For each species there are 6 months of the annual cycle in which any given climate variable positively influences flowering intensity and 6 months of negative influence. For all species, rainfall exerts a negative influence when temperature is positive.