452 S.-Afr.Tydskr. Plantk. , 1989, 55(4): 452-463 Grass genera of southern Africa: Interactive identification and information retrieval from an automated data bank L. Watson*, G.E. Gibbs Russell 1 and M.J. Dallwitz Taxonomy Laboratory, Research School of Biological Sciences, Australian National University, GPO Box 475, A.C.T. 2601, Australia; Botanical Research Institute, Private Bag X1 01, Pretoria, 0001 Republic of South Africa and Division of Entomology, Commonwealth Scientific and Industrial Research Organization, GPO Box 1700, Canberra City, A.C.T. 2601, Australia 1 Present address: Missouri Botanical Garden, P.O. Box 299, St. Louis, Missouri, USA 63166-0299 A new interactive system for computerized taxonomic identification and information retrieval, operable on MS­ DOS microcomputers, is now available. In this application to southern African grass genera the data include synonyms, morphology, anatomy, physiology, cytology, classification, pathogens, world and local distribu­ tions and references. The program is part of the DELTA system, which also provides natural language descriptions and printed keys. The data for the southern African genera can be provided separately, or inclu­ ded in the complete data for the grass genera of the world. 'n Nuwe interaktiewe stelsel vir gerekenariseerde taksonomiese identifisering en inligtingsherwinning, bedryf­ baar op MS-DOS-mikrorekenaars, is nou beskikbaar. Hierdie toepassing op Suider-Afrikaanse grasgenusse bevat data oor sinonieme, morfologie, anatomie, fisiologie, sitologie, klassifikasie, patogene, wereldwye en plaaslike verspreidings en verwysings. Die program is deel van die DELTA-stelsel, wat ook voorsiening maak vir beskrywings in gewone taal en gedrukte sleutels. Die data vir die Suider-Afrikaanse genusse kan afsonderlik verskaf word, of by die volledige data vir die grasgenusse van die wereld ingesluit word. Keywords: DELTA, expert systems, identification, information retrieval, INTKEY, Poaceae *To whom correspondence should be addressed Introduction to be fully realized , it will be necessary to reconsider the We here announce the availability and illustrate the objectives, and consequently the practical organization, capabilities of a new computerized system for interactive of descriptive taxonomy. Tn particular, the samples of identification and information retrieval for grasses, with interrogations reproduced below show that if special emphasis on applications in a southern African automation is limited to the sort of data found in conven­ context. In doing so, we will demonstrate some of the tional printed floras , a major advance will have been advantages of automating taxonomic descriptions. The made (see Examples 1- 3) . However, aciting oppor­ applicability of automation at different levels of the tunities for automation lie beyond traditional floristics , taxonomic hierarchy, and its ability to cope with diverse in linking taxonomic data to results from allied organisms and very large data sets have been discussed disciplines, thus extending fundamental taxonomic and exemplified by Watson et al. (1988) , with reference research , applying taxonomic expertise in othe r biolog­ to successful applications to all genera of caesalpinoid ical disciplines, and perhaps attracting much-needed legumes and grasses, to floristic descriptions of the funds into systematics (see Examples 4-7). genera, species, subspecies and varieties of Paniceae in Australia, as well as to beetles and plant viruses. The Procedures application described here comprises grass genera, and The DELTA system foreshadows an automated system for the complete grass The DELTA system (Dallwitz 1980; Dallwitz & Paine flora of southern Africa, preparation of which is already 1986) was designed to satisfy a need (cf. Watson & Milne under way. This example also illustrates for the first time 1972) for a generalized system for accommodating all the a new interactive program which is a significant exten­ different kinds of descriptive data used by taxonomists, sion of the facilities available in the DELTA system. without information loss, in an easy-to-use format Computerization of taxonomic descriptions offers designed to minimize coding errors. An associated great advantages over traditional methods. New data can program, CONFOR, translates the coded descriptions be incorporated as they accrue, so that descriptions, into natural language, offering a choice of formats ; classifications and keys can be continuously updated; produces summarized data for specified sets of taxa, and the range and flexibility of the new identificatory giving for multistate characters the numbers of taxa and information retrieval techniques made available by exhibiting each character state and for numeric charac­ automation bring within easy reach operations which ters giving the means, ranges and names of taxa were previously impossible or hopelessly laborious. We representing the extremes of ranges; and carries out are on the threshold of a new era in taxonomic descrip­ various data maintenance operations - for example, tive work, which will incorporate a significant automated changing the sequences of characters and character component. It is worthwhile taking this opportunity, states while automatically keeping all the files consistent therefore , to show that if the potential of automation is with one another. Data coded in DELTA format can S.Afr.J. Bot. , 1989, 55(4) 453 Example 1 Identify a fragmentary specimen from Using the keywordfor southern Africa. eliminate irrelevant taxa and southern Africa (detailed provenance unknown), limited explore the available vegetative characters. to vegetative parts and inflorescence remains with empty glumes. Start by setting an appropriate MATCH Enter command: INCLUDE TAXA SAFRICA command, and obtaining a list of character and taxon 203 taxa included. keywords. Enter command: CHARACTERS VEGETATIVE 3: <longevity of plants> 4: <habit> Enter command: MATCH IOU 12: culms <whether branched above> MATCH set to Inapplicables, Unknowns, Overlap. 38: leaf blades <mature: foldedlrolled> Enter command: KEYWORDS CHARACTERS 46: adaxial ligule <form - avoid seedlings> CHARACTER KEYWORDS USed AVailable Use three of the 47 available vegetative characters. ALL Nomenclature Enter command: 4 Vegetative form 4: <habit> HAbit 1. long-rhizomatous CUlms (form) 2. long-stoloniferous LEAVes (form) 3. caespitose <Figs 1,7> REProductive organization 4. decumbent <including 'rooting at the nodes'> <Fig. 2> INFlorescence form Enter value: 2 FEMSterile spikelets 81 taxa remain. FEMFertile spikelets GLumes Enter command: 12 INComplete florets 12: culms <whether branched above> FLorets (female-fertile) 1. branching <vegetatively> above <Fig. 2> LEMmas (female-fertile) 2. unbranched <vegetatively> above <Figs 1,7> AWns of female-fertile lemmas Enter value: 2 PALeas (female-fertile) 68 taxa remain. ANdroecium of female-fertile florets GYnoecium Enter command: 38 FRuit 38: leaf blades <mature: foldedlrolled> SEedling 1. flat LEAF blade anatomy 2. folded <Fig. 42> LB epidermis 3. rolled PHOtosynthetic pathway, etc 4. acicular ULtrastructure of the leaf blade Enter value: 2 Biochemistry 30 taxa remain. TS Anatomy of the leaf blade TSCulm Search for ligule characters by word, obviating the need to scan the DIAgnostic features of individual taxa list for them. CYtology CLassification Enter command: CHAR "ligule" SUBfamilies 45: adaxial ligule <presence> SUPertribes 46: adaxial ligule <form - avoid seedlings> TRibes and subtribes 47: adaxial ligule <shape of apex> SPecies number 48: adaxial ligule <length at middle: generally recorded only for Ecology membranous, un fringed forms> Weeds 49: abaxial <outer> ligule <data very incomplete> GEography PHYtogeographical Kingdoms, Regions, etc Enter command: 46 DIStribution 46: adaxial ligule <form - avoid seedlings> REFerences 1. an unfringed membrane <may be variously hairy or ciliolate> DNa <Fig. 19> PAThogens 2. a fringed membrane <Figs 20,21,23> RUsts 3. a fringe of hairs <Fig. 22> SMuts 4. a rim of minute papillae HYbrids Enter value: 2 COmments 17 taxa remain. PReset for diagnostic descriptions Search for observable inflorescence characters. Enter command: KEYWORDS TAXA Enter command: CHAR INFLOR TAXON KEYWORDS 62: inflorescence <whether possessing pseudospikelets: see Remaining McClure 1973 for defmition> Eliminated ALL 70: inflorescence <whether digitate or subdigitate: state 1 requires AUstralia subdivision> TEAfrica 71: with <number of primary inflorescence branches: applied mainly Safrica to forms with spike-like main branches - data very incomplete> Greece 454 S.-Afr.Tydskr. Plantk., \989 , 55(4) 100: spikelets <whether in regular long-and-short combinations, as Obtain a description ofCynodonfor the characters used, and check exemplified in typical andropogonoids> the specimen against it. Enter command: DESCRIBE REMAINING Enter command: 70 Enter character ranges or keywords: USED 70: inflorescence <whether digitate or subdigitate: state 1 requires 172. Cynodon <Rich.> subdivision> 4: 1. digitate or subdigitate <includes paired branches> <Figs 38, 1. long-rhizomatous 54> 2. long-stoloniferous 2. not digitate <neither digitate nor 'subdigitate' - implicit> 38: leaf blades Enter value: 1 1. flat 5 taxa remain. 2. folded 136. Chloris 46: adaxial ligule 160. Craspedorhachis 2. a fringed membrane 172. Cynodon 3. a fringe of hairs 506. Paspalum 70: inflorescence 690. Tetrapogon 1. digitate or subdigitate 71: with Enter command: 71 2-20 primary inflorescence branches 71: