Morphological variability of ( L, Quercus petraea (Matt) Liebl, Quercus pubescens Willd) in northeastern France: preliminary results Jl Dupouey, V Badeau

To cite this version:

Jl Dupouey, V Badeau. Morphological variability of oaks (Quercus robur L, Quercus petraea (Matt) Liebl, Quercus pubescens Willd) in northeastern France: preliminary results. Annales des sciences forestières, INRA/EDP Sciences, 1993, 50 (Suppl1), pp.35s-40s. ￿hal-00882873￿

HAL Id: hal-00882873 https://hal.archives-ouvertes.fr/hal-00882873 Submitted on 1 Jan 1993

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Original article

Morphological variability of oaks (Quercus robur L, Quercus petraea (Matt) Liebl, Quercus pubescens Willd) in northeastern France: preliminary results

JL Dupouey V Badeau

Laboratoire de phytoécologie forestière, Centre de recherches forestières, INRA Nancy, 54280 Champenoux, France

Summary — Morphological variability of oaks in Lorraine (northeastern France), was studied. Eight hundred oaks were sampled in 80 stands covering a broad range of ecological variability; 10 , fruits and current-year shoots were collected per . Thirty-four morphological variables were measured and analyzed by factorial correspondance analysis. It is concluded that Q robur and Q pe- traea are clearly separated with a few morphologically intermediate individuals (3.5%). Q petraea is more variable than Q robur. Q pubescens and Q robur are totally isolated from each other, while Q petraea and Q pubescens form a continuum. Many variables discriminate between these 3 species; some of them have been little known prior to now (pilosity, presence of intercalary ribs). These re- sults are compared with those from other parts of Europe. / morphometrics / hybridization / introgression / Quercus robur / Quercus petraea / Quercus pubescens

Résumé — Variabilité morphologique des chênes dans le Nord-Est de la France; résultats préliminaires. Nous avons étudié la différenciation morphologique des chênes pédonculé, sessile et pubescent dans le Nord-Est de la France. L’échantillonnage a porté sur 80 populations provenant de stations représentant toute la gamme de variation des milieux de chênaies en Lorraine. Sur 10 arbres par population, 10 feuilles, infruiescences et rameaux de l’année ont été prélevés. Trente- quatre variables morphologiques ont été mesurées et analysées par analyse factorielle des corres- pondances. On observe une très nette séparation des chênes sessile et pédonculé, avec seulement 3,5% d’individus morphologiquement intermédiaires, ainsi qu’un isolement total du chêne pubescent et du chêne pédonculé. Par contre, les chênes sessile et pubescent forment un continuum. Le chêne pédonculé est moins variable que le chêne sessile. De nombreuses variables discriminent ces 3 espèces, dont certaines peu connues jusqu’alors (pilosité, présence de nervures intercalaires). Ces résultats sont comparés à ceux obtenus par ailleurs en Europe. taxonomie / morphométrie / hybridation / introgression / Quercus robur / Quercus petraea / Quercus pubescens INTRODUCTION total of 761 were sampled during summer 1989, 655 of them had produced fruit during this year. Twenty leaves, fruits (including peduncles, The distinction of species in the Quercus cupules and acorns) and twigs of the current complex is still a matter of debate. In west- growth year were collected. To minimize posi- tional within the tree and Jen- ern Europe, several species have been re- variability (Blue sen, 1988), leaves were collected from the ex- ported as potentially interbreeding, the ternal part of the canopy usually on the aspect most Quercus robur and widespread being facing south, and always in the middle part of Quercus petraea. Until now, the prevalent the first flush shoot. Ten of these 20 samples opinion was in favor of the common occur- were chosen at random for measurements, after rence of hybrids between the different spe- the elimination of broken, incomplete or dam- cies, producing many morphologically in- aged units. Eighty variables were measured or calculated. These variables concern as- termediate forms between pure parental many pects of foliar and fruit morphology: size, overall species due to hybridization and introgres- shape, color, pilosity on various parts of leaves sion. A of literature has been huge body or fruits (measured as in Grandjean and Sigaud, published with this thinking in mind, (see 1987), shape of some details (lobes, auricles at eg, Kissling, 1983; Minihan and Rushton, the lamina base). These data were acquired 1984, for the most recent papers). Only a with a digitizing tablet hooked up to a microcom- few recent studies (Dupouey, 1983; Du- puter. Several features were obtained from dis- tances and between different landmarks pouey et Le Bouler, 1989; Dupouey et al, angles along the outline of the . 1990; Grandjean and Sigaud, 1987; Aas, Data were factorial corre- 1990) have come to different conclusions. analyzed mainly by spondence analysis using the SPAD.N statisti- The aim of this study was to describe cal package (Lebart et al, 1988). In the first the actual morphological status of Q robur stage of the analysis, the study of correlation co- and Q petraea at a regional scale (Lor- efficients between all pairs of the 80 initial vari- raine Plain), including Q pubescens. We ables allowed the elimination of 46 redundant studied inter- and intraspecific variations, variables. The comparison between results with or without fruit morphological characters showed and their link with constraints. In ecological no significant differences so only a subset of 29 this results on mor- preliminary paper, only parameters describing leaves and shoots was phological differentiation at the interspecif- used for subsequent analysis. This allowed the ic level are presented. We have tried to use of the whole set of trees instead of only answer the following questions: what is the those which had fruited during the sampling organization of morphological variability year. Fruit variables were used as supplemen- tary characters. among the 3 species? What is the degree Variables were ranked of isolation of each species? What are the by decreasing power of discrimination to their F value in an best discriminant morphological charac- according unbalanced analysis of variance between the 3 ters? species. Bonferroni t-tests of difference between means for each species were performed. Dis- criminant analysis was used to calculate a func- MATERIALS AND METHODS tion for species recognition.

Stands were selected from 8 forests in the Lor- RESULTS raine Plain, and a total of 80 populations were sampled in order to cover the whole ecological variability of oakwoods in this area. Coppices 1 shows the of trees into with standards were discarded, as were sup- Figure projection posed plantations. In each of these populations, the space of axes, 1, 2 and 3 of the factori- 10 dominant trees were marked at random. A al analysis. One can observe 3 poles of distribution for these individuals. Table I At the left end of the first axis, and at gives the mean values of the most discrim- the upper part of axis 2, leaves have a inant variables for each pole. long petiole. The maximum width is at the At the right-hand side of the first axis, middle of the lamina. Fruits are shortly pe- dunculate and is medium to dense. the leaves are shortly petiolated, with well- pilosity developed auricles at the base of the this pole is composed of Q petraea trees. lamina. The maximum width of leaves is The last pole is also located on the left- located in the upper part of the lamina. hand side of the first axis, in the lower part Lobe sinuses are irrigated by numerous in- of axis 2. It is composed of trees with tercalary veins. Fruits have a long and thin leaves rather similar to those of the previ- peduncle. Pilosity is absent, very short on ous one. Lobes are sharper, often accom- all parts. This group of individuals repre- panied by lobules (lobes irrigated by sents the typical Q robur. nerves of the third order). The pilosity is more highly developed, both in terms of specific morphological variabililty than Q density and length. This group can be robur, which appears to be more homoge- identified as Q pubescens. neous. Thus we find, with this analysis, that the Table I gives the variables by decreas- 3 species have different morphological ing power of discrimination between the poles, and also that these 3 species differ species. The best discriminant parameters widely in their degree of separation from are pilosity (density and length of pilosity one another. The Q robur cluster is com- on nerves, petiole and lamina), number pletely separated from that of Q pubes- and length of intercalary veins, length of cens, and only a few morphologically inter- the petiole and of the peduncle. Some mediate individuals are found between Q classical features only appear after these robur and Q petraea (3.5% of the total variables, such as the development of an number of trees). auricle at the lamina base. On the other hand, Q petraea and Q pu- Length of the petiole and number of in- bescens form a continuum without any ercalary veins are sufficient to separate the clear distinction between the 2 species. A 2 species Q robur an Q petraea with 99% number of morphologically intermediate in- success rate. The discriminant function for dividuals occurs. Consequently, Q petraea tree recognition is: and Q pubescens exhibit much more intra- where, nint. number of intercalary veins analysis. The main methods used in these (mean of 10 leaves per tree); lpet: length cases were hybridity indices for which you of petiole in mm (mean of 10 leaves per had to choose subjectively the range of tree); I species index; positive for Q robur each species for all the variables before and negative for Q petraea. Values be- calculation, principal components analysis tween -1000 and +1000 indicate an inter- (PCA) using ’pure’ stands as references mediate tree. and discriminant function analysis. On the other hand, purely descriptive methods, such as factorial correspondence analysis DISCUSSION AND CONCLUSION (or principal components without reference populations) do not require the definition of the species before the analysis. They can These results are different from those re- be regarded just as a means of looking at in a number of ported previous publica- the raw data from a particular point of view tions: Carlisle and Brown (1965), Wigston (the one with maximum variance ex- (1974), Olsson in Sweden (1975 a,b) plained). Rushton (1978, 1979, 1983), Minihan and Factorial is Rushton (1984) in the United Kingdom and correspondence analysis to PCA when a ’reference Kissling (1980a,b, 1983) in Switzerland, all preferable (even is not used) because it is able concluded that there was extensive hybrid- population’ to deal with non-linear be- ization between Q robur and Q petraea. relationships tween characters, whereas PCA only Conversely, the authors of some more re- measures linear correlation coefficients. cent studies reached the same conclu- sions we did: Grandjean and Sigaud The frequency of hybrids between Q pu- (1987) in France (including Q pubescens), bescens and Q petraea has been under- letswaart and Feij (1989) in The Nether- lined by other authors. Semerikov et al, lands; Aas (1990) in Germany. (1988), studying populations from Dage- stan (Russia), even conclude that such hy- It is difficult to compare these results be- brids a cause are based on different represent unique species complex. they sample This lack of isolation could the sizes, sampling regimes, characters and explain greater variability observed in Q petraea with different companion species. But sev- versus Q robur. eral hypotheses could be advanced to ex- plain the discrepancies. One possible ex- Variables used in previous taxonomic planation is that the extent of hybridization studies were not always the most highly differs in the different parts of the distribu- discriminating ones, and sometimes not in tion range of the species. Peripheral situa- discriminant at all. The best features, tions would be more favorable to hybridiza- our sample, are intercalary veins and pilos- tion than in the inner part. This could ity development. account for results from Sweden or the Further studies in progress are explor- United Kingdom, but not those from Swit- ing the persistence of these discriminant zerland. More probably, one must look to characters under homogeneous growing the differences in the statistical approach- conditions (nursery) for the populations un- es used for data analysis. All previous der consideration. Also, the structure of the studies concluding that numerous hybrids intraspecific variability along ecological which were present were based on statisti- clines is of major importance. Finally, there cal methods according to which the limits is a need for standardization of the statisti- of each species are defined before the cal methods used for the analysis of mor- phological variability in the different parts and Quercus petraea in The Netherlands. 313-325 of the range of distribution of these spe- Acta Bot Neerl 38, cies. Kissling P (1980a) Un réseau de corrélations en- tre les chênes (Quercus) du Jura. Ber Schweiz Bot Ges 90, 1-28 ACKNOWLEDGMENTS Kissling P (1980b) Clef de détermination des chênes médioeuropéens (Quercus L). Ber Schweiz Bot Ges 90, 29-44 We gratefully acknowledge the technical assis- Kissling P (1983) Les chênaies du Jura central tance of Gilles Doucet during the collection of suisse. Thesis, Université de Lausanne, Swit- leaves. We also thank Patrick Behr for meas- zerland, 400 p urements of fruits and Hervé Cochard for 3D Lebart L, Morineau A, Lambert T for this was (1988) graphics. Support project provided SPAD.N: Portable Pour the EEC action Système l’Analyse by (Research Programme des Données. Version 1.2: Manuel de Référ- MA1B: Genetics and Breeding of Oaks). ence. CISIA, Paris, 306 pp Minihan VB, Rushton BS (1984) The taxonomic status of oaks in Breen Wood, REFERENCES (Quercus ssp) Co Antrim, Northern Ireland. Watsonia 15, 27-32 Aas G (1990) Kreuzbarkeit und Unterscheidung Olsson U (1975a) A morphological analysis of von Stiel-und Traubeneiche. Allg Forstztg 9- phenotypes in populations of Quercus (Faga- 10, 219-221 ceae) in Sweden. Bot Not 128, 55-68 Blue MP, Jensen RJ (1988) Positional and sea- Olsson U (1975b) On the size and microstruc- sonal variation in (Quercus, ) ture of pollen grains of Quercus robur and leaf morphology. Am J Bot 75, 939-947 Quercus petraea. Bot Not 128, 256-264 Carlisle A, Brown HF (1965) The assessment of Rushton BS (1978) Quercus robur L and Quer- the taxonomic status of mixed oak popula- cus petraea (Matt) Liebl: a multivariate ap- tions (Quercus sp). Watsonia 6, 120-127 proach to the hybrid problem. 1. Data acqui- Dupouey JL (1983) Analyse multivariable de sition, analysis and interpretation. Watsonia quelques caractères morphologiques de pop- 12, 81-101 ulations de chênes (Quercus petraea (Matt) Rushton BS (1979) Quercus robur L and Quer- Liebl et Quercus robur L) du Hurepoix. Ann cus petraea (Matt) Liebl: a multivariate ap- Sci For 40, 265-282 proach to the hybrid problem. 2. The geo- Dupouey JL, Le Bouler H (1989) Discrimination graphical distribution of population types. 209-224 morphologique des glands de chênes ses- Watsonia 12, sile (Quercus petraea (Matt) Liebl) et pe- Rushton BS (1983) An analysis of variation of donculé (Quercus robur L). Ann Sci For 46, leaf characters in Quercus robur L and Quer- 187-194 cus petraea (Matt) Liebl population samples Dupouey JL, Fougère V, Kremer A (1990) Vari- from Northern Ireland. Ir For 40, 52-77 abilité génétique des chênes sessile et pé- Semerikov LF, Glotov NV, Zhivotovskii LA donculé. Rev For Fr42, 198-203 (1988) Example of effectiveness of analysis Grandjean G, Sigaud P (1987) Contribution à la of the generalized variance of traits in trees. taxonomie et à l’écologie des chênes du Ber- Sov J Ecol 18, 140-143 ry. Ann Sci For 44, 35-66 Wigston DL (1974) Cytology and genetics of letswaart JH, Feij AE (1989) A multivariate anal- oaks. In: The British Oak (Morris MG, Perring ysis of introgression between Quercus robur FH, eds). Bot Soc Br Isles, London, 27-50