Ann. For. Sci. 59 (2002) 777-787
DOI: 10.1051/forest:2002065

Leaf morphological differentiation between Quercus robur and Quercus petraea is stable across western European mixed oak stands

Antoine Kremer^a, Jean Luc Dupouey^b, J. Douglas Deans^c, Joan Cottrell^d, Ulrike Csaikl^e, Reiner Finkeldey^f, Santiago Espinel^g, Jan Jensen^h, Jochen Kleinschmitⁱ, Barbara Van Dam^a0, Alexis Ducousso^a, Ian Forrest^d, U. Lopez de Heredia^aa, Andrew J. Lowe^c, Marcela Tutkova^e, Robert C. Munro^c, Sabine Steinhoffⁱ and Vincent Badeau<sup>b

a</sup>  Institut National de la Recherche Agronomique (INRA), Unité de Recherches Forestières, BP 45, 33611 Gazinet Cedex, France
^b  Institut National de la Recherche Agronomique (INRA), Unité d'Écophysiologie Forestière, 54280 Champenoux, France
^c  Centre for Ecology and Hydrology, Edinburgh, (CEH), Bush Estate, Penicuik, Midlothian, EH26 0QB, Scotland, U.K.
^d  Forestry Commission (FC), Forest Research, Northern Research Station, Roslin, Midlothian, EH25 9SY, Scotland, U.K.
^e  Austrian Research Centre (ARCS), Seibersdorf, 2444, Austria
^f  Swiss Federal Research Institute WSL, Zürcherstrasse 111, Birmensdorf, Switzerland
^g  NEIKER A.B., Granja Modelo-Arkaute, 01080 Vitoria-Gasteiz, Spain
^h  Danish Forest and Landscape Research Institute (DFLRI), Hørsholm Kongevej, Hørsholm 2970, Denmark
ⁱ  Niedersächische Forstliche Versuschsanstalt Fortamstr. 6, 34355 Staufenberg-Escherode, Germany
^a0  ALTERRA, Green World Research Institute, Wageningen, The Netherlands
^aa  ETSI Agrarias, Universidad de Valladolid, Avda Madrid 57, 34004 Palencia, Spain

(Received 2 April 2001; accepted 8 November 2001)

Abstract
Leaf morphology was assessed in nine mixed oak stands (Quercus petraea and Q. robur ) located in eight European countries. Exhaustive sampling was used in an area of each stand where the two species coexisted in approximately equal proportions (about 170 trees/species/stand). Fourteen leaf characters were assessed on each of 5 to10 leaves collected from the upper part of each tree. Three multivariate statistical techniques (CDA, canonical discriminant analysis; PCA, principal component analysis; MCA, multiple correspondence analysis) were used in two different ways: first on the total set of leaves over all stands (global analysis) and second, separately within each stand (local analysis). There was a general agreement of the results among the statistical methods used and between the analyses conducted (global and local). The first synthetic variable derived by each multivariate analysis exhibited a clear and sharp bimodal distribution, with overlapping in the central part. The two modes were interpreted as the two species, and the overlapping region was interpreted as an area where the within-species variations were superimposed. There was no discontinuity in the distribution or no visible evidence of a third mode which would have indicated the existence of a third population composed of trees with intermediate morphologies. Based on petiole length and number of intercalary veins, an "easy to use" discriminant function applicable to a major part of the natural distribution of the species was constructed. Validation on an independent set of trees provided a 98% rate of correct identification. The results were interpreted in the light of earlier reports about extensive hybridization occurring in mixed oak stands. Maternal effects on morphological characters, as well as a lower frequency or fitness of hybrids in comparison with parent species could explain the maintenance of two modes, which might be composed of either pure species or pure species and introgressed forms.

Résumé
La différenciation morphologique des feuilles entre Quercus robur et Quercus petraea est stable à travers les peuplements mélangés de chênes de l'ouest européen. La variabilité de la morphologie foliaire a été etudiée dans neuf peuplements mélangés de chênes (Quercus petraea et Q. robur) en Europe sur la base d'un échantillon exhaustif moyen de 170 arbres/espèce/peuplement. Trois méthodes d'analyses multivariables ont été utilisées (ACD : analyse canonique discriminante ; ACP : analyse en composantes principales ; AFC : analyse factorielle des correspondances). Les trois méthodes aboutissent à des résultats congruents. La première variable synthétique de chaque méthode se caractérise par une distribution bimodale, chaque mode correspondant à une espèce. La distribution de la première variable ne manisfeste pas de discontiunuité pouvant indiquer l'existence d'un troisième groupe, correspondant à des arbres avec des morphologies de type intermédiaire. Une méthode d'identification rapide basée sur les deux caractères les plus discriminants (nombre de veines intercalaires et longueur du pétiole) a été proposée et validée sur un autre jeu de données (98 % d'identifications correctes). La structure de la variation morphologique entre les deux espèces résulte sans doute de l'hybridation entre elles, et de l'hérédité des caractères morphologiques.

Key words: morphology / leaf / Quercus robur / Quercus petraea / taxonomy

Mots clés : morphologie / feuille / Quercus robur / Quercus petraea / taxonomie

Correspondence and reprints: Antoine Kremer Tel: +33 5 57 97 90 74; fax: +33 5 57 97 90 88; antoine.kremer@pierroton.inra.fr

© INRA, EDP Sciences 2002