Techniques for Reconstructing Data on Flowering

• Published in: Russian journal of ecology Vol. 34; no. 1; pp. 59 - 61
• Main Authors: naciari, M, Orlandi, F, Cenci, C A, Romano, B
• Format: Journal Article
• Language: English
• Published: New York Springer Nature B.V 01.01.2003
• ISSN: 1067-4136; 1608-3334
• DOI: 10.1023/A:1021871223059

One of the objectives of applied phenological investigations is to reconstruct, for various species, historical series of data on the same biological phenomenon of interest, such as budding, receme formation, floral bud development, the beginning of flowering, full flowering, the end of flowering, fruit setting, and the development of fruits (Efron, 1994; Liu, 1995; Heitjan, 1997). Phenology offers a good tool for studying plants and, in particular, interactions between climate variation and plant behavior (Chuine, 1999). However, in order to use conventional statistical techniques, complete series of phenological data collected over many years are necessary. The purpose of this work was to test different methods for reconstructing missing data and to compare their reliability. In the past, the missing data were usually substituted by the average value of the recorded data. Alternatively, the data on species with the same periods of flowering were used (Bricchi et al., 1995), and the results obtained by this method were good, even though more variable (Bassi et al., 1996; Cenci et al., 1997). We compared three techniques for reconstructing missing data--smoothing, similar split-plot design, and correlation technique--using the results of 20-year phenological observations on 50 plant species.