- Title
- Wavelets and clustering: methods to assess synchronization
- Creator
- Hudson, Irene L.; Keatley, Marie R.; Kang, In
- Relation
- Wavelets: Classification, Theory and Applications p. 97-124
- Relation
- Mathematics Research Developments
- Relation
- https://www.novapublishers.com/catalog/product_info.php?products_id=28629
- Publisher
- Nova Science Publishers
- Resource Type
- book chapter
- Date
- 2012
- Description
- In this chapter the discrete wavelet transform (DWT) is shown to accurately enumerate and visually demonstrate synchronisation of a flowering time series. The DWT is applied to flowering intensity records (c~vering 1940 - 1970) of Eucalyptus leucoxylon, E. microcarpa, E. polyanthemos and E. tricarpa. The DWT is used as a correlational method, and proposed as a new metric for synchronisation and visualisation for a multidimensional time series. The wavelet based results are compared to Moran's traditional synchronisation test and to a self organising map time series clustering approach. The Maximal Overlap DWT analysis successfully identify four subcomponents in each flowering series characterised as a (i) non-flowering phase (dl), (ii) duration (d2), (iii) annual (d3) and (iv) intensity (d4) cycles. The wavelet correlation (WCORR) signatures are shown to exhibit a common pattern for synchronous species pairs, for example d2, d3 and d4 WCORR values being significant and positive for E. leucoxylon and E. tricarpa whereas for species pairs that are asynchronous the d2 and d3 specific WCORR values are significant and negative for E. tricarpa and E. polyanthemos. The profiles of wavelet cross correlation (WCCORR) enable us to further understand how synchrony may relate to the subcomponents of flowering (peak annual cycles, start, cessation, duration, overlap) and climate. The positive periods of the WCCORR profiles for d2 reflect flowering overlap between species pairs. The d3 WCCORR's peak at the lag which is the difference between the species pair's peak flowering months. Distinctly different patterns for d3 WCCORR are associated with asynchronous and synchronous time series. The sinusoidal and cycling of the d3 specific WCCORR profiles mirror the cyclic 6 monthly climatic influence on species flowering, where for 6 months if temperature is positively correlated with flowering, rainfall is negative and vice versa.
- Subject
- discrete wavelet transform; wavelets; clustering; synchronisation
- Identifier
- http://hdl.handle.net/1959.13/1335764
- Identifier
- uon:27487
- Identifier
- ISBN:9781621002529
- Language
- eng
- Hits: 1575
- Visitors: 1648
- Downloads: 0
| Thumbnail | File | Description | Size | Format |
|---|