Kost H., Rudakova V., Volkova S., Shipunov A.
Abstract
Two Russian and Ukrainian ivy species: Hedera helix and H. colchica were investigated for clarifying which metric leaf characters are most suitable for distinguishing these species in fieldwork conditions. Multivariate visualisation methods (like principal component analysis and classification trees) show that most perspective characters are lengths of the leaf ``diagonal'', petiole and internode. Crimean (``Hedera taurica'') populations do not differ significantly from Caucasian H. helix.
Introduction
Two ivy (Hedera L.) species are growing on the South Russia and Ukraine: Hedera helix L. and H. colhica (C. Koch) C. Koch, sometimes H. taurica Carr. for Crimean plants is also accepted (Rubtsov, 1972). These species are usually hardly distinguished from each other. The situation relates with high variability of leaf shapes and sizes, which commonly used for determination. Another characters to distinguish ivy species are the number of chromosomes (H. helix 2n = 48, while in H. colhica 2n = 192 - Vargas et al. 1999), smell (pungent and nutmeg-like, respectively - Zernov 2002) and the structure of leaf trichomes (stellate or plate-like, respectively - Lum & Maze 1989; Ackerfield 2001; Ackerfield & Wen 2002). Unfortunately, all mentioned characters are more or less unsuitable for fieldwork. So the goal of our investigation was to find some optimal metric characters which can be used for separating this species during field observations.
Materials and methods
Material collected in Krasnodar region of Russia (Tuapse and Big Sochi districts) and Crimea. These regions are only part of European Russia and Ukraine with warm temperate weather conditions. The work was carried out in spring 2002 and 2003. We have 8 localities investigated in Caucasus and 20 localities in Crimea. Totally 346 leaves were measured. Most of characters (except structure of trichomes) were measured in field.
Ivy has at least three types of shoots: (1) vegetative, ground-based; (2) vegetative, trunk-based and (3) generative. Shoots with long and short internodes also exist, but we used leaves only from long-internode shoots. On each shoot leaves changed their size from shoot base upwards periodically: at the beginning of each ``period'' they are small, than became bigger and then smaller again. Such period we called ``series'' according with N. Krenke (1933) point of view. We measured all leaves in one (usually top) series from several shoots in each locality.
The following characters were observed: (1) TIP.POB: shoot type (0 - ground-based; 1 - trunk-based; 2 - generative); (2) L.MEZHD: length of an internode (from the base of petiole), mm; (3) L.CHER: length of a petiole, mm; (4) W.LOC: position of a greatest width of a leaf, mm; (5) DIAG: length of the ``diagonal'' of a leaf i.e. the distance from the leaf base point to the right apical cut, mm; (6) L.MAX: highest length of a leaf, mm; (7) W.MAX: highest width of a leaf, mm; (8) PARTS: number of the leaf lobes; (9) SMELL: smell type from pulverised leaves (pungent or nutmeg-like) and (10) TRICH: structure of the trichomes (stellate or plate-like). Besides, we calculated some ``secondary'' characters which utilise measured ones: (a) L.MAX/L.CHER = relative length of petiole; (b) L.MAX/W.MAX = leaf ``circularity''; (c) W.LOC/L.MAX = leaf ``oblongness'' and (d) DIAG/(sqrt(2)*W.MAX) = diagonal fraction, i.e. fraction between ``theoretical diagonal'' (supposing that leaf is round-shape) and real one. Thus we have different group of characters: (i) primary metric characters (L.MEZD, L.CHER, W.LOC, DIAG, L.MAX, W.MAX); (ii) categorical ones (SMELL, PARTS, TRICH) and (iii) secondary metric characters.
Then analysis of this multivariate data was performed using statistical software R (Venables & Smith 2002).
Results and discussion
All primary leaf-based characters, petiole length and internode length are correlated to each other (Spearman r = 0.5-0.8; p < 0.05). PARTS also correlates with most of leaf-based characters (r ( 0.5; p < 0.05). Unfortunately, SMELL is strongly associated with person who was performed measurements and hereby this character is very subjective. Further investigations showed that most people could recognise pungent smell of H. helix, but nutmeg-like smell of H. colchica is very hard to distinguish. All leaf characters had low correlations (r < 0.4; p < 0.05) with trichome type. All secondary characters more or less correlate with each other and with primary leaf characters. They also have much less variation than primary characters.
To explore the overall diversity of measured leaves we performed several principal component analyses (PCA). First PCA was performed with primary metric characters, PARTS and TRICH. Resulted classification is relatively clear and generally accords with distribution of TRICH character. However, this classification does not correspond with distribution of PARTS. This means that whereas classification from first PCA does not contradict with primary metric characters, PARTS does not play a big role here. Resulted classification does not depend from shoot type and all leaves from one series are placed together. Crimean populations do not separate from Caucasian H. helix, so there is an argument against separate species status for Crimean ivy. Analysis of principal components' loadings showed that biggest loadings in this case have DIAG, L.MAX and L.MEZHD.
For proving the role of trichome type character we removed it and than repeat principal component analysis with remained characters. This second PCA, however, cannot return clear groups.
Third PCA was performed with the set of secondary characters. The resulting classification is not clear and there are no groups. The TRICH character has much less loading than in first PCA. This can mean that ``classification behaviour'' of secondary characters are disagree with TRIH and primary leaf characters and thereby not suitable for classification.
Discriminant analysis was also performed to elucidate relationships between classification by metric characters and trichome type. Last two characters were used as classifying variables. Percent of correctness (number of correct classifications / number of objects * 100) in this case is relatively big (88.4%). Then we used a recursive partitioning analysis (Venables & Ripley 2002) to clarify which metric characters are most useful for classification. This analysis usually returns so-called regression tree that corresponds with contributions of all characters in classification model. Our tree shows that DIAG character plays a biggest role in classification: most of leaves (92%) that have DIAG less than 25.5 mm is leaves with plate-scale trichomes. Consequently, some of metric characters (especially DIAG, L.CHER and L.MEZHD) could be used for diagnostic purposes.
Conclusions
I. Lengths of the leaf ``diagonal'', petiole and internode are most perspective characters for the field determination of Hedera colchica and H. helix.
II. Crimean (``Hedera taurica'') populations do not differ from Caucasian H. helix.
III. Caucasian localities usually contain both groups (Hedera helix and H. colchica). Unfortunately, there is no clear evidence of coexistence, hybridisation or introgression between this species.
IV. Number of leaf lobes does not correspond with classification by trichome types.
V. Classification by trihome types does not distinguish the different types of the shoots.
We acknowledge Dr. S. Glagolev, A. Kvashenko and students of Moscow South-West High School and Dr. A. Oskolski (Botanical Institute, Saint-Petersburg) for their great help in this work.
References
Ackerfield, J. (2001): Trichome morphology in Hedera (Araliaceae). - Edinb. J. Bot. 58: 259-267.
Ackerfield, J. & Wen, J. (2002): A morphometric analysis of Hedera L. (the ivy genus, Araliaceae) and its taxonomic implications. - Adansonia. 24: 197-212.
Krenke N. P. (1933): Fenogeneticheskaja ismenchivost'. (Phenogenetic variability.) - Trudy Biol. instituta im. K. A.Timirjazeva. 1. [In Russian]
Lum, C. & Maze, J. (1989): A multivariate analysis of the trichomes of Hedera L. - Watsonia. 17: 409-418.
Rubtsov N. I. (1972): Opredelitel' vysshikh rastenii Kryma. (Diagnostic book of Crimean higher plants.) - Leningrad : Academic Publishers. [In Russian]
Vargas, P. & McAllister, H. A. & Morton, C. & Jury, S. L. & Wilkinson, M. (1999): Polyploid speciation in Hedera (Araliaceae): phylogenetic and biogeographic insights based on chromosome counts and ITS sequences. - Pl. Syst. Evol. 219: 165-179.
Venables, W. N. & Ripley, B. D. (2002): Modern Applied Statistics with S. New York: Springer Verlag.
Venables, W. N. & Smith, D. N. & the R Development Core Team (2002): An Introduction to R. Bristol: Network Theory Ltd.
Zernov, A. S. (2002): Opredelitel' rastenij Severo-Zapadnogo Zakavkaz'ja. (Diagnostic book of Northwest Trans-Caucasus plants.) - Moscow: KMK Scientific Press.