Muelleria 24: 111-136(2006) The taxonomic status of Lachnagrostis scabra, L. aequata and other related grasses in Australia (Poaceae). Austin J Brown Primary Industries Research Victoria, Department of Primary Industries, 621 Sneydes Road, Werribee. Abstract Early descriptions and treatments of Lachnagrostis scabra (Beauv.) Nees ex. Steud.and L. aequata (Nees) S.W.L. Jacobs are examined and compared to current determinations of herbaria collections. The results of this study show a confusing array of nomenclature and inconsistency in descriptions given of these taxa. Additional assessment of a range of morphological characters on more than 100 specimens find coastal L. scabra and L. aequata to be inseparable, but, taxonomically distinct from collections previously determined as L. scabra from inland Tasmania. The name L. scabra takes precedence for the coastal entity while the intend material is published as a new taxon, L. morrisii A.J. Brown and shown to be distinct from L. collicola (D. Morris) S.W.L. Jacobs. Further related and new taxa, L. scabra subsp. curviseta A.J. Brown from Victoria and L. nesomvtica A.J. Brown subsp. nesomytica. L. nesomytica subsp. pseudofiliformis A.J. Brown and L. nesomvtica subsp. paralia A.J. Brown from Western Australia are also described. Introduction There has long been uncertainty about the taxonomic relationship between Lachnagrostis scabra (syn. Agrostis rudis) and L. aequata (A. aequata). Early treatments had very few specimens on which to form judgements but with many collections being made over the last 50 years or so, the opportunity to more rigorously assess this situation is now possible. The current paper reports on a re-examination of the earliest published descriptions and on some detailed moiphological assessment of the bulk of extant collections. This paper is part of an ongoing series dealing with a reassessment of the genus for Australia. Note that unless indicated as n.v. all cited collections have been viewed by the author. In her revision of Australian species of Agrostis, Vickci^ (1941) separated A. aequata Nees from A. rudis Room. & Schult., based on the degree of scabridity of the sides of the glumes. She differentiated the former from the latter “in the smooth sides of the glumes, which arc less distinctly acute, equal or subequal both in length and breadth, not exceeding 2 mm in length”. However, she also “retained both species until further material and field studies enable a conclusion to be reached". In her paper, she only cites three collections of each taxon, including the types; all but one, originating from Tasmania. The exception was an A. aequata collection from Nelson in Victoria. All the collections dated from 1792 to 1894 and had coastal locations (where provided). TAXONOMIC TREATMENT IN STATE FLORAS POST VICKERY (1941) Black (1943) tentatively recorded A. aequata for South Australia on the basis of the Nelson collection. Jessop and Toelken (1986) added A. rudis to the South Australian 1 12 Brown flora, based on a few specimens from the South-east region. They also noted that A. aequata had been collected twice from the extreme south of the South-east. In Victoria, Willis (1970) recorded A. ntdis from a number of localities in southern Victoria; Main Ck near Arthurs Seat, Lake Corangamite, Gorae West near Portland and the Lower Glenelg R. from Moleside Ck downstream to Nelson and noted that collections from Lake Corangamite and Nelson had been determined as A. aequata by Vickery. Vickery (1941) had not previously cited the Lake Corangamite collection, despite it being one of Mueller’s from 1875. Willis considered A. aequata to be “a slight variant" of A. ntdis; its differing characters being “features merely of degree”. Walsh and Entwisle (1994) agreed with Willis, regarding A. aequata as “being easily accommodated within the range of variation exhibited by A. ntdis". Jacobs and Pickard (1981) recorded A. aequata for the Southern Tablelands of NSW, based on a single collection from 1949. Curtis and Morris (1994) recognised both A. nulis and /(. aequata for Tasmania but noted the former as a plant of inland areas (“damp areas on margins of forest from 200 to 1200 m altitude”) and limited the latter to coastal habitats. They reported that A. aequata “has generally smaller spikelets, glumes usually smooth and panicles in which the majority of spikelets have unawned lemmas.” Jacobs (2001a) placed A. ntdis and A. aequata in Lachnagrostis as L. ntdis (Roem. & Schult.) Trin. and L. aequata (Nees) S.W.L. Jacobs, respectively. Jacobs (2002) later corrected L. ntdis to L. scabra (Beauv.) Nees ex Steudel. Recently, Barker et al. (2005), in their census ol vascular plants ol South Australia, accepted Lachnagrostis as a valid genus and regarded L. aequata as a synonym of L. scabra. EARLY DESCRIPTION AND NOMENCLATURE REVIEW The type for Agrostis scabra R.Br. non Willd., presumably collected by Robert Brown, is without location, collection number or date, although a notation on the reverse side of the collection sheet states “Van Diemens Land - Adventure Bay”. Adventure Bay is located on the east side of Bruny Island in SE Tasmania. Adventure Bay was also the location cited by Vickery (1941) for a collection “in swamps” of A. ntdis by Nelson in I-ebruary 1792 (n.v.). However, as David Nelson died of a inflammatory fever in Timor on 20.7.1789, after enduring 41 days at sea in a small open boat with Captain William Bligh and 17 other men, following the Mutiny on the Bounty (Orchard 1999), either the date or the collector or both must be incorrect. Prior to the mutiny. Nelson had made botanical collections during the Bounty's anchorage at Adventure Bay (Orchard 1999) from 20th August to 4th September 1788, but this late winter-early spring period is not conducive for the collection of mature Lachnagrostis. Bligh did return to Adventure Bay on the Providence from 9th - 24lh February 1792 but with James Wiles and Christopher Smith as botantists (Orchard 1999). Further collections were made at that time, the bulk of which were eventually lodged at BM with duplicates at DPU, FI, G, HO and K and it seems more likely that the ‘Nelson’ collection derives from this material. Unfortunately, the Adventure Bay collection cannot be located and does not appear to have been viewed since Vickery did so in 1938. However, the handwriting on the type collection for Agrostisparvijlora RBr. from Adventure Bay “in black vegetable soil” and also labelled as Nelson, February 1792 in fact has been recently identified as that of Wiles (pers.com. R. Vickery). Nelson’s name has been added to the sheet later by an unknown hand, perhaps in the belief that he was still Bligh’s botanist in 1792. It is assumed that the A. ntdis from Adventure Bay was treated likewise. Lachnagrostis 113 Robert Brown's trip to Australia of 1802-05 began as botanist for Captain Matthew Flinders on the Investigator. During the circumnavigation of Australia, Brown did make plant collections on King Island in Bass Strait (Moore 2000) and in Port Phillip Bay (Willis 1966) but only very late in the season (from April 23lli to May 3ld 1802). There is no mention of any grass collections in his diary notes of this period (Vallance et al. 2001). After the Investigator was condemned as unseaworthy. Brown returned to Tasmania on the Lady Nelson, leaving his collecting trips in the vicinity of Port Jackson on 28lh November 1803. Originally planning to sail to Port Phillip, the Lady Nelson reached the Kent Island group in Bass Strait on 11December where plant (mainly algae) collections were made (Moore 2000, Vallance et al. 2001). January lsl to 18'1’ was spent at Port Dalrymple and along the Tamar River to the South Esk Gorge, near to the present site of Launceston. It is in this period that Brown records Agrostis along with other grasses on Green Island (opposite Outer Cove - now York Cove) on 3ui January 1804 and “immediately above the western arm” (a little further south of Green Island) on 5th January. Nine days were spent in Port Phillip Bay, Victoria from 21s1 to 30'1' January, assisting in packing up the ill-fated. Lt Col David Collins led settlement at Sullivans Bay (present day Sorrento) but Brown kept no diary of this period (Vallance et al. 2001). He then sailed south to Risdon Cove and Sullivans Cove, near present day Hobart, Tasmania with the settlers from Port Phillip Bay. From 9lh February to 9,h August 1803, after which he finally left Tasmania, Brown spent his time in the SE, largely exploring the Derwent River, Huon River and Mt. Wellington region. He made one late autumn-winter sea voyage towards the SW, including anchorage at Adventure Bay on 12,h June (Vallance 2001). This appears to be the only occasion when Brown visited this site. In Brown (1810), the description of his A. scabra collection is annotated with “J.D.”, the 'J' referring to “vicinity of Port Jackson (including Hunter’s River, Coal River)” and the ‘D' to “Van Diemens Land” respectively. Bentham (1878) only refers to Brown’s collection as deriving from “Port Jackson”, while Vickery (1941) indicates “without locality”. Given the lack of any other collections from coastal NSW. the Port Jackson location is highly unlikely. The reference to Adventure Bay on the reverse side of the sheet also appears erroneous; given the maturity of the collection but the winter landing on Bruny Island. The notation may refer to the earlier ‘Nelson’ (actually Wiles) location instead or in fact, the specimen may be a duplicate taken from that collection. If Brown did collect his own specimen, the most likely location would be at Port Dalrymple during early January 1804 or possibly at Port Phillip Bay in late January 1804. Although an undated collection oi' Agrostis aentula (syn. Lachnagrostis aetmila) was made at Port Dalrymple by Col Patterson during the following summer and passed on to Brown in early 1805 (Vickery 1941. Vallance et al. 2001), L. scabra was also collected from “Tamar” in March 1887 by Oakden. The island, shoreline and creek mouth habitat where Brown observed his Agrostis would be conducive to L. scabra growth whereas the forested hillsides around Port Dalrymple would be more likely habitat for/., aemula. Brown’s (1810) description of his A. scabra, though brief, included keeled glumes with scabrid sides and a glabrous lemma with a denticulate apex and a spikelet- contained, dorsal awn. Roemer and Schultes (1817) and Steudel (1855) adopted Brown’s description without change. Meanwhile, Nees (1843), when describing A. aequata, based on an 1838 collection by Gunn (1005) from Van Diemen’s Land, describes the lemma as papery, smooth and obtuse and the glumes as equal with scabrid keels. There is no mention of the presence or otherwise of an awn. When the two taxa were illustrated by Fitch and described by Hooker (1859) for Tasmania, the representations of his “A. scabra Br." are clearly not the same as Brown’s 114 Brown collection but appear to be a species or variant o(Deyeiixia. The illustration (Plate CLXb) is rather similar to Fitch’s rendering of A. contractu F. Muell. ex I look.f. (syn. D. contractu (F. Muell. ex Hook.f.) Vickery), except that the lemma has a short, near apex awn and the palea is shorter, while Hooker's text includes D. fiigida Mueller mss. (syn. D.frigida F. Muell. ex Benth.) as a synonym of .4. scabra and suggested that the taxon may have been a variant of A. montana R.Br. non Krock (syn. D. montiebla (Roem. & Schult.) Vickery). Hooker describes the taxon's spikelets as being “rough, often purple, /6 inch long”. At over 4 mm, these spikelets are somewhat larger than the 2.5 mm glumes, described by Vickery (1941) for A. rudis. Hooker also describes the lower palea (i.e. lemma) of A. scabra as “very rough, hard, shaip, nerves indistinct, awnless, or with a short, hardly exserted awn inserted at about the middle of the back”. Apart from the inconsistent insertion of the awn, Fitch's illustration of the floret depicts Hooker’s description well. Hooker describes the habitat of A. scabra as being subalpine; from the foot of Mount Wellington, Arthurs Lakes etc.; citing collections from Gunn (1461, 1462), Archer and Oldfield (without numbers). In her treatment of Deyeiixia, Vickery (1940) recognised that Fitch’s illustration and Hooker’s description of A. scabra R.Br. were incorrect and assigns them to D. scabenda Vickery, although Gunn's “Arthurs Lakes” sample and Oldfield’s unnumbered collection suggests that Hooker also included D. accedens Vickery in his A. scabra. Vickery (1940) placed the Tasmanian specimens of D. frigida cited by Bentham (1878) (Oldfield; New Norfolk, Gunn) in D. accedens. Despite her conclusions about Fitch and Hooker's rendering of A. scabra, Vickery (1941), when revising Agrostis still assigned Hooker’s description to A. rudis. This was undoubtedly an oversight on her behalf. Hooker’s (1859) description of A. aequata spikelets expands on Nees (1843) account by stating that the glumes are “ovate, acute, hardly longer than the flower, very scabrid, especially along the keel” and that the paleas are “rather coriaceous, with obscure nerves, awnless; lower (i.e. lemma) truncate; upper (i.e. palea in the modern sense) nearly as long, with two small teeth at the tip, and sometimes a small tuft of hairs at the base” (“basi extus pedicello gracili” i.e. rachilla extension). Fitch’s illustration also shows the glumes and lemma apex with ciliate margins. The collection by Gunn (1005) is cited by Hooker (1859) and is likely an isotype (Gunn, No. 1005/1837, 18.1.1838, H2003/00I80 K.) as Hooker’s Plate No. CLIXb and pencil sketches of Fitch’s illustration appear on one of the two sheets on which the type is mounted. Also attached to this type is a comment by Vickery, made on 9.7.1938 that it “matches type of Agrostis aequata Nees at Cambridge”. Bentham (1878) cited a few collections of these grasses from Tasmania but created a very confused scenario by his treatment of them. Gunn’s collection (1005) was transferred to Deveuxia as D. aequata Benth. (syn. A. aequata Nees), while he placed Brown's A. scabra and a collection from Adventure Bay (probably the erroneous ‘Nelson’), in D. scabra Benth. However, also included in D. scabra were D. contractu and D. decipiens (R.Br.) Vickery (syn. A. decipiens R.Br.). Vickery (1940) regarded Bcntham’s description of D. scabra as also including her D. scabenda. Bentham (1878) also designated the North American species, A. scabra Willd., for specimens from Tasmania, Victoria and NSW (regarding A. parviflora R.Br. as part of the species) and therefore, as a consequence, his appears to be the first Australian treatment to recognise A. scabra R.Br. non Willd. (i.e. as Deyeiixia scabra Benth.) as a separate taxon. The North American entity has, in the past, been regarded as a synonym for A. hiemalis (Walt.) Britton, Sterns and Poggenb. but has most recently been regarded as a separate taxon (Soreng and Peterson 2003). Although Vickery (1941) Lachnagrostis 115 “with some hesitation” placed a range of Australian specimens, including at least some of Bentham’s citations (excluding A. parviflora), in A. hiemalis, Jacobs (2001b) has subsequently replaced it with two new endemic species; A. propinqua S.W.L. Jacobs and A, bettyae S.W.L. Jacobs.. In Tasmania, Rodway (1903) followed Bentham’s (1878) treatment for Deyeuxia and Agrostis but included Circular Head (probably in reference to Gunn’s collection) and Trial Harbour (collection by Rodway, December 1894) as locations for D. aequata. His illustration of D. scabra Bentli. is likely to be of D. scaberula while that for A. scabra Willd. appears to be of A. parviflora. For his Victorian plant key, Mueller (1888) combined all Deyeuxia and Agrostis under Agrostis; listing A. scabra (probably as a synonym for D. scabra Benth.) and A. rudis (presumably for A. scabra R.Br. non Willd.). The latter would appear to have been made on the basis of the collection he made from Lake Corangamite in 1875. Tate (1890) recorded A. scabra Willd. for the Mount Gambier district of South Australia; probably on the basis of a collection of A. scabra R.Br. non Willd. he made in 1882 from Nelson, in Victoria, near the South Australian border (Tate Herbarium. 15.11.1882) and without taking account of Bentham’s (1878) separation of the North American taxon from the Australian entity or the presence of paleas in the florets of his specimen. Black (1922) recognised but renamed Tate’s collection, Calamagrostis aequata (Nees) J.M. Black for South Australia but Ewart (1930) seems to have overlooked these taxa altogether for Victoria. His Calamagrostis rudis Steud. is cited as “a mountain grass” and probably refers to a Deyeuxia species (possibly D. contracta) while his A. scabra refers to the North American taxon. Materials and Methods MORPHOLOGICAL DESCRIPTION At least 95% (134 specimens) of all current collections, determined as L. scabra and L. aequata at HO, MEL, AD and NSW were examined and measurements taken of leaf width, pedicel length (shortest and longest), glumes (lower and upper), glume setae, lemma, lemma setae, awn, palea, anther and rachilla extension length (including hairs) and the height of awn (if present) attachment on the lemma baCk Ratings of the degree of scabridity of the keels (1-3) and sides (0-2) of the glumes and cilia of the glume margins (0-2) and lemma apex nerves (0-2) were made. Measurements of spikelet and floret characters were made on one ‘typical’ mature example for each specimen. Panicle exsertion from the leaf sheath was noted and, if observed, peduncle length, panicle height and width and the distance between the two lower-most branch whorls (whorl separation) were measured. In addition to types for L. scabra and L. aequata lodged at BM and K respectively, the 134 examined specimens (including the occasional duplicate) were as follows; Tasmania: 29 specimens of L. aequata from 23 locations, 23 specimens of inland L. scabra from 20 locations, 8 specimens from 3 sheets of L. collicolci from 3 locations; Bass Strait Islands: 14 specimens of L. aequata from 8 locations; Victoria: 26 specimens of L. scabra from 21 locations; South Australia: 12 specimens ol L. scabra from 10 locations; Western Australia: 21 specimens of an indeterminate taxon from 14 locations on Rottnest and Garden Islands Among the Tasmanian collections, one determined as L. aequata (Archer undated and without location) was actually akin to 116 Brown inland L. scabra while the reverse was true of another collection (Buchanan 8.1.1987 Wallaby Bay, Port Davey). In addition to the collections as outlined above, two specimens from NSW, determined as L. scabra (originally determined as A. aequata) were also examined. One collected from Eden was found to be Deyeuxia parviseta var. boormanii. The other NSW collection, from Cathcart in 1949, was a very mature specimen with few intact spikelets and consequently difficult to determine with certainty. It appeared to conform to L. scabra in most discernible characters and could not be likened to any other taxonomic candidate, despite it’s somewhat disjunct location from the known distribution range of the species. Further fresh collections from the field arc required before confirmation can be assured. During examination of specimens determined as L. scabra at HO (hereafter referred to as ‘inland L. scabra'), some similarities with the Tasmanian montane species, L. collicola (D. Morris) S.W.L. Jacobs were observed. Subsequently, morphological characters for L. collicola, were compared to inland L. scabra to test their relationship. In the course of this study, a collection at MEL of five sheets by Walcott in 1881 from Rottnest Island, Western Australia of an indeterminate taxon of Lachnagrostis were examined. Labelled by Mueller as Agrostis sp. it has obvious affinities with the L. scabra/L. aequata material. Subsequently, more recent indeterminate PERTH material from Rottnest Island (one specimen) and Garden Island (three specimens) were examined, measured, and compared to the first. An additional seven collections were made by the author on Rottnest Island in December 2004 and, although very mature, were still intact enough to examine vegetative, inflorescence and spikelet characteristics. A further nine collections were made in November 2005, at a time when plants were still green and inflorescences were emerging and becoming divergent. STATISTICAL TREATMENT Except for L. collicola, which was treated as a separate entity, examined specimens were grouped according to geographical source; i) Tasmania Mainland - Inland, ii) Tasmania Mainland - Coastal, iii) Tasmania - Bass Strait Islands, iv) East Coast Victoria and New South Wales, v) West Coast Victoria and South Australia and vi) Western Australia. A range of statistical analyses were undertaken using Genstat 8.1 (Release PL16, Lawes Agricultural Trust, Rothamstead Experimental Station). A correlation matrix (CORRELATE) for the character measurements was produced. In addition, a Symmetric Matrix (SYMMETRICMATRIX) of associations for the characters was established by forming a Similarity Matrix (FSIMILAR1TY) (using euclidean variate type for continuous variables and cityblock variate type for discontinuous variables). This matrix was used in both a Principle Coordinates Analysis (PCO) and a Hierarchical Cluster Analysis (HCLUSTER) (using Furthest Neighbour criterion) to assess natural groupings of the specimens. All of the measured characters were used in the multivariate analysis except for inflorescence height, width and distance between lower branch whorls (excluded because of unmeasurable characters in many specimens due to non-exserted panicles). Analysis of Variance (ANOVA) (LSD at 5%) for each measured character, was also performed using i) geographical source and ii) 75% threshold cluster groups, as treatments. For awn and inflorescence measurements, ANOVA was restricted to specimens with awns or exserted panicles respectively. Lachnagrostis 117 Table 1: Means in morphological characters (all lengths in mm) and percent of specimens with exserted inflorescnces, hairy lemmas and awned lemmas for various taxa (designated by both old and new specific names) and geographical groupings (values followed by the same letter are not significantly different at 95% confidence limits). Character Tas. Tas. Tas. Bass East Vic West Vic West Aust Mainland Mainland Mainland Islands and NSW and SA Islands locale inland inland coastal coastal coastal coastal island old name colUcola scabra a e if u at a aequata scabra scabra indet. new name coUicola morrisii scabra scabra scabra scabra nesomytica Leaf width 1.6 ab 2.4 cel 1.9 be 1.5 ab 2.1 bed 2.7 d 1.2 a InJlor. assertion* 38 61 66 86 60 75 81 Inflor. height1 77 a 109 ab 142 ab 135 ab 132 ab 174 A 109 aft Inflor. width1 33 a 104 ab 89 ab 104 ab 105 ab 111 A 112 6 Whorl separation1 14 a 24 ab 38 abc 26 ab 39 abc 44 be 64 be Shortest pedicel 13.3(7 11.3 c 2.1 a 1.6 a 1.9 a 1.9 a 5.9 6 Longest pedicel 28.5 c 27.4 c 7.1 a 7.6 a 7.3 a 6.2 a 17.6 ft Lower glume 3.81 e 2.56 cl 2.22 c 1.49 a 1.90 b 1.95 be 2.54 (7 Glume setae 0.18c 0.08 h 0.07 b 0.00 a 0.02 a 0.03 ab 0.01 a Upper glume 3.76 cl 2.54 c 2.20 be 1.54a 1.91 b 1.90 ft 2.33 c Keel scabridsb 2.0 b 2.0 h 2.6 c 2.5 c 1.9ft 2.0 ft 1.3a Lateral scabrids1’ 1.8 c 0.7 b 1.1 b 1.4 c 1.5 c 1.3 Ac 0.1 a Glume margins11 0.6 A 0.1 a 1.9 cl 2.0 cl 1.9(7 1.9(7 1.0 c Lemma length 3.11(7 2.11 c 1.69 b 1.30 a 1.68 ft 1.61 A 1.62 ft Lemma setae 0.33 c 0.06 ab 0.09 b 0.01 a 0.03 ab 0.03 ab 0.01 a Lemma nerve1’ 0.0 a 0.0 a 1.8c 1.5 Tic 1.6 c 1.1 ft 0.0 a Hairv lemmas'' 0 0 0 0 0 25 38 Lemma hairsb: - - - - - 0.1 a 1.0 A Callus hairs 0.49 c 0.33 b 0.25 ab 0.11 a 0.30 A (cid:9632) 0.33 ft 0.22 aft Awn presencea WO 100 38 29 33 5-1 100 Awn attach3 89.6(7 94.1 cl 71.1 ab 75.6 be 79.5 be 76.5 Ac 63.5 a Awn length3 2.50 6 0.61 a 0.51 a 0.36 a 0.48 a 0.38 a 1.74 A Palea length 2.51(7 1.83 c 1.61 b 1.26 a 1.55 ft 1.52 ft 1.49 ft Rachilla extens. 2.14(7 0.97 be 0.79 b 0.46 a 1.01 Ac 1.10c 0.96 Ac Anther length 0.70(7 0.50 be 0.53 c 0.41 a 0.41 a 0.45 aft 0.52 be a Percent of specimens with character, b discrete variable recorded as a rating, 1 Restricted to specimens with exserted inflorescences.2 Restricted to specimens with hairs on lemmas,3 Restricted to specimens with lemma awns Results MORPHOLOGICAL ASSESSMENT The correlation matrix (not provided) showed high (r2 = 0.85-98) positive relationships between spikelet (glumes) and floret (lemma and palea) lengths as would be expected. In relation to glume and lemma lengths, anther length was moderately correlated (0.62- 0.70), rachilla extension length was less correlated (r2 = 0.55-0.64) and awn length was poorly correlated (0.32-0.59). Over all taxa, awn attachment height was poorly correlated (r = - 0.39) to awn length although, within the WA island populations, correlations were high and negative (r2 = - 0.86). High negative correlations (r2 = - 0.77-0.82) were also found between glume margin cilia and the length of the shortest and longest pedicels, and pedicel lengths were moderately and positively correlated (r2 = 0.64-0.67) to glume length. In other words, spikelets with longer pedicels had less evident cilia and tended to be larger. Glume margin cilia was positively correlated (r2 = 0.69) to lemma apex nerve cilia but glume (lower and upper) scabridity (keels and laterals) were poorly correlated (0.27-0.53) to these characters. Other correlations (e.g. involving leaf width, callus hair length, inflorescence size) were mostly low (r < 0.50). Figures I and 2 show the results of PCO and hierarchical cluster analyses respectively. 118 Brown Figure 1. Vectors 1 and 2 of principle coordinate analysis of all examined specimens; L. collicola - Tasmania (• ) ,inland L. scabra - Tasmania (+), L. sp. - Western Australian islands (A), L. aequata - Tasmania ((cid:9633)), L. aequata- Bass Islands ((cid:9632)), L. scabra - Western Victoria/South Australia (0), L. scabra - Eastern Victoria ((cid:9830)). indeterminate L. sp. - Cathcart, NSW ( A). Lachnagrostis collicola, inland L. scabra and the WA islands plants are clearly separated in the PCO analysis from each other and strongly separated from a widely scattered and intermingled grouping containing coastal L. scabra and L. aequata (Figure 1). Vectors 1 and 2 of the PCO accounted for 51.3% of the total variation with vectors 3, 4 and 5 providing an extra 7.5%, 7.2% and 6.1% respectively. Table 1 shows the means in results for the characters measured for the various taxa and geographical groupings. Lachnagrostis collicola is significantly different from the other taxa in having larger spikelets (lower and upper glume and glume setae length) and florets (lemma, lemma setae, awn, palea, callus hairs, rachilla extension and anther lengths). Ranges in both glumes, lemma, palea and rachilla extension length do not overlap with the ranges for these measurements on the other taxa. Lachnagrostis collicola is similar to inland L. scabra in always being awned and having a similar and high awn attachment, having Lachnagrostis 119 70 65 60 55 50 -45 40 35 coastal L scabra/L aequata + S + E L sp. (VVA) + C inland L. scabra _ A. collicola _ Figure 2. Dendrogram of hierarchical cluster analysis (furtherest-neighbour criteria) for all specimens (C = Cathcart specimen, S = Sherbrooke River specimen, E = Eight Mile Creek specimen); at 75% separation. the same range in pedicel length and in lacking ciliate lemma nerves. However, in addition to its larger spikelets and florets, it differs from the latter, in almost always having narrower leaves and more scabrid sides to its glumes. The inland L. scabra from Tasmania is hereby recognised as a new taxon, separate to both L. collicola and coastal L. scabra/L. aequata and is described as L. morrisii A. J. Brown. The material from Rottnest and Garden Islands is most similar to the montane L. morrisii but has, on average, smaller lemmas and paleas, longer awns that have lower lemma attachment, shorter callus hairs (often absent), less scabrid glumes but with more marginal cilia, shorter pedicels and narrower leaves with distinctly separate ranges for lemma length and awn attachment. From L. scabra/L. aequata, the specimens differ in having, on average, longer glumes (except for the Tasmanian group), awns and pedicels, but particularly in having less scabrid and/or ciliate glumes and lemma apices. These character differences, its habitat and its geographical isolation from the eastern Australian taxa have prompted the recognition of this entity as a new species; described as L. nesomytica A.J. Brown. Although all the collections of L. nesomytica are superficially similar and were not clearly separated in the PCO (at least for the first two vectors) or hierarchial cluster analyses, some populations, as well as being geographically separated, are distinct enough to warrant recognition as separate subspecies. The populations on Rottnest Island are confined to the edges of saline lakes and swamps and obviously have high salt tolerance. Populations at Bickley Swamp, Serpentine and Government House Lakes (from the south-east part of the Island), differ from the remaining populations by having geniculate awns and moderately hairy lemmas. In these characters they resemble a stunted form of L. filiformis (Forst.) Trin. but have smaller spikelets and florets, longer anthers, shorter awns, longer pedicels and stiffer inflorescences with more divaricate branches. Typical L. filiformis has not been collected from the Island, although it occurs reasonably commonly on the adjacent mainland. The hairy lemma form of L. nesomytica is described here as subsp. pseudojiliformis A..I. Brown. The populations of L. nesomytica from the nearby Garden Island have been collected from coastal dunes in thickets of Callistris preissii and Melaleuca lanceolata; a highly calcareous but less saline environment (McArthur 1957) than that encountered by the other subspecies. Although having glabrous lemmas, as for typical L. nesomytica, they have longer, often curved to geniculate awns; though these are not as long as for subsp. pseudojiliformis. This entity is described as subsp. paralia. The distributions of L. aequata from coastal Tasmania and the Bass Strait Islands and coastal L. scabra from the mainland are not separable and readily encompass the types for both taxa (Figure 1). They arc therefore regarded as a single taxon; L. scabra 120 Brown (Beauv.) Nees ex Steudel. Table I shows some significant differences in characters between geographical location for the taxon (e.g. the Bass Is collection has the smallest spikelets and florets and the Tasmanian collection has the largest anthers), but overall ranges (not provided) in these characters overlap. The degree of scabridity of the sides of the glumes, which w'as the main character used by Vickery (1941) to distinguish between Agrostis mdis and A. aequala, was found to be extremely variable and poorly correlated with any other measured character (the best correlation being only r2 = 0.32 with the degree of glume margin ciliation). The presence of lemma awns is not linked to geographical location or correlated to spikelet/floret size. In some cases, both awned and unawned lemmas appear on the same specimen. Inequality in glume lengths bears little relationship to geographical location or to any other measured character. Most specimens of L. scabra from the Fumeaux Group of Bass Strait Islands, were small tussocks with divergent inflorescence branches and awnless or minutely awned lemmas. Apart from the general lack of awns, these plants have some resemblance to L. nesomytica subsp. nesomytica. However the latter have significantly longer pedicels and lack the cilia of lemma apices and the degree of glume scabridity. Although it is tempting to separate the Bass Strait Islands L. scabra from the mainland entity, there does not appear to be any clear difference in morphology, apart from tussock and inflorescence size. Plants of very' similar tussock and inflorescence appearance but with hairy and long-awned lemmas are also found on these islands and are likely to represent a stunted form of L.filiformis. Another similar entity with hairy lemmas and very short, non-geniculate awns also occurs on Flinders Island, but its relationship to other Island forms is still under investigation by the author. Examination of a previously indeterminate specimen at AD, collected by Eardley in 1942 from Eight Mile Creek, near Port MacDonnell, SA, was found to conform to L. scabra in all respects except that some of its lemmas had scattered short hairs near the margins of the lateral surfaces. Despite Eardley's comment that the grass ‘comes in everywhere there’, recent field searches for this population by the author has failed to find it. Closer examination of other specimens of L. scabra found at least five others with lemma hairs, all from South-east SA to South-west Victoria. However, these appeared to have only the occasional short hair on the occasional lemma and are therefore easily overlooked. Other specimens could well have the same condition. The sporadic and seemingly haphazard occurrence of this trait provides little support for creating a taxonomic distinction within L. scabra. Another specimen, at MEL, labelled as L. aff. scabra, from Sherbrooke River, Port Campbell National Park (Beauglehole and Finke 21182, 6.ix. 1966), also has an occasional hair on lemma margins but is distinguished from other L. scabra collections, in having longer (up to 1.4 mm) and often curved to somewhat flexuose or geniculate, very fine awns. This character appears distinctive enough to designate the specimen as a new taxon; L. scabra subsp. curviseta. The specimen from Wallaby Bay. Port Davey {Buchanan 9339, 8.i. 1987), that had previously been determined at HO as inland L. scabra, has some awns from 0.9-1.1 mm long but these are straight and not curved.