An attempt towards a natural system of Harpullia (Sapindaceae)

Starting point of the present study on Harpullia was the taxonomic revision of that genus by Leenhouts & Vente (Blumea 28, 1982, 1-51). The system to which that revision led is primarily intuitive and accordingly subjective. The intention of the present paper is to give a more natural system based upon more objective criteria and with the use of more scientific methods. The fundamentals of a systematic study are species and characters (see chapter 2). The species are those presented in the taxonomic revision; the characters used are tabulated in the form of a synoptic key (see chapter 2 B). In chapter 3 a survey is given of two intuitive systems, the one by Leenhouts & Vente and the preceding one by Radlkofer (1933-34). The first approach was towards a phenetic system (chapter 4). The method used is a kind of simplified numerical taxonomy with weighted characters. Weighting of characters is based upon the supposition that a character which is constant in most of the taxa concerned is heavier than one that varies in several taxa. The phenetic system to which this led is expressed in figure 1 and given at the end of chapter 4. In chapter 5 a phylogenetic approach is given. Phylogenetic valuation of characters, primarily with the use of out-group analysis, secondary of correlation, is discussed (5 A). All species got a phylogenetic formula, giving the primitive or derived states of the characters used, and a phylogenetic value expressing the degree of primitiveness or derivative (5 B). The method used is a simplified Hennigian cladistics, resulting in a kind of Wagner tree (fig. 2). In chapter 6A a comparison is made between the intuitive systems, mainly the one by Leenhouts & Vente, the phenetic and the phylogenetic system developed here, and the pollen morphological phylogenetic system given by J. Muller (Blumea 31, 1985, 161-218, this issue). As a whole there appears to be a good agreement in many points between the different systems. The main exceptions are Harpullia cupanioides and still more so H. hillii because of their variability, H. longipetala that is macromorphologically rather primitive but palynologically advanced, and H. rhachiptera with a very aberrant pollen type and an uncertain position in the phylogenetic system. A translation of phylogenetic systematics into formal taxonomy has been tried in chapter 6 B. It appeared difficult to express the branching of the phylogenetic scheme in the hierarchy of taxonomy, mainly because taxa should be delimited by clear boundaries defined by good, preferably conspicuous characters. The subdivision of Harpullia into two subgenera, Otonychium, incl. H. arborea and H. pendula, and Harpullia with the further 24 species is indisputable. More hesitatingly, a further division of subg. Harpullia into two sections, Harpulliastrum with only H. austro-caledonica and Harpullia with the other species is accepted. For different reasons it appeared impossible or undesirable to split off some other at first sight well circumscribed groups, however.