MADRO?O, Vol. 50, No. 4, pp. 300-306, 2003 SEQUENTIALLY ADJUSTED SEX-RATIOS IN GYNOMONOECISM, AND POA DIABOLI (POACEAE), A NEW SPECIES FROM CALIFORNIA ROBERT .1. SORENG Department of Systematic Biology?Botany, National Museum of Natural History, Smithsonian Institution, Washington DC 20013-7012 soreng.rob@nmnh.si.edu DAVID J. KEIL Biological Sciences Department, California Polytechnic State University, San Luis Obispo, CA 93407 ABSTRACT Poa diaboli is described as a new and rare species off. snbsect. h'ladropoa from tiie coastal mountains of southern California. Its breeding system is like that in species of Poa previously said to be partially gynodioecious. However, in species o? Poa with the corresponding breeding system syndrome, plants that are totally pistillate-?owered are rare and possibly only late-flowering individuals. We conclude that the breeding system of these species of Poa is better classified as a new subset of gynomonoecism, here termed sequentially adjusted gynomonoecism. Subtypes of gynomonoecism in the grasses, and in Poa, are outlined. Key Words: breeding systems, Gramineae, grass, gynomonoecism, Poa, Taxonomy. The new species of Poa L. was possibly first col- lected by David Keil in 1998. Because of its un- certain identity, and isolated geographical occur- rence from similar species, it was brought to the attention of first auth?r It was initially thought to be dioecious, but further investigation of new col- lections with longer series of specimens, revealed that it is either weakly gynodioecious or exhibits an unusual form of gynomonoecism. Poa (subsect. Madropoa) diaboli Soreng & Keil, sp. nov. (Fig. 1)--TYPE: USA, California, San Luis Obispo Co., W end of San Luis Range, Montana de Oro State Park, slopes W of Valen- cia Peak, very local on steep N facing slope in coastal scrub, 164-245 m, 27 March 1998, D. Keil 26474 (holotype: US, isotypes: OBI-55333, [CAS, K, RSA; to be distributed]) A Poa confine robustiore, altiore (26-50 cm versus 7-30 cm), paniculis longioribits (4-8.5 [10.5] versus 7-5 cm), diffusioribus, laminis planis vet plicatis, lemmatis longioribus ([3.2?] 4.25?5 versus 2.4-4 mm), venis sine pilis, racliillis vi- sibilis longioribus plerumque asperate sine pilis, plantis gynomonoeciis, antheris plerumque lon- gioribus ([1.4] 1.75-2.6 versus 1.5-2 mm), dif- fert. Sequentially adjusted gynomonoecious, perenni- al grass, rhizomatous and stoloniferous, IbtTiiing loose, leafy, tufts up to 25 cm tall and 30 cm across. Vegetative shoots numerous, shoot emergence ex- travaginal with rudimentary prophylls, and pseudo- intravaginal with tubular prophylls, and intravagi- nal with well developed prophylls. Culms 26-50 cm long, 0.5-0.9 mm diam., numerous, widely spreading to sortly decumbent and erect, leafy, one or more nodes exposed, uppermost node in lower third to middle of culm, frequently branching above the base with several shoots developing in from ad- jacent nodes with contracted internodes. Leaves; cataphylls present on extravaginal shoots; sheaths keeled, lightly scabrous, uppermost 4.5?9 cm long, margins fused -/??''/?o the length; collars scabrous to pubescent on margins; ligules of middle and upper culin leaves 1.0?3.0 min long (of lateral shoots and basal culm leaves less than 1 mm long), apex trun- cate and lacerate to (mostly) obtuse or acute, api- cally scabrous, abaxial surface moderately to densely scabrous; blades flat (especially upper culm ones) or more often folded, thin, slender, 0.8-2.0 (?2.5) mm wide, narrowly prow-tipped, abaxial surface smooth, veins prominent, keel and margins scabrous, adaxial surface of innovation blades sparsely to moderately densely scabrous or short- hairy on and between the veins (infrequently gla- brous between veins), culm blades generally grad- ually decreasing in length upwards, uppermost blades 2.9-6 (-7) (pistillate specimens), or (3-) 4- 8.6 (-11) (perfect specimens) cm long; upper- sheath to blade-length ratio 1.4-2.4: 1, averaging 1.8 (pistillate specimens), 0.6-2.1: 1, averaging 1.4 (perfect specimens). Paniculate synflorescences ex- erted above the vegetative mounds, erect to nod- ding, pyramidal to ovate, open, or eventually loose- ly contracted, sparse, 4.0-8.5 (-10.5) cm long; branches 1?2 per node, spreading to ascending, slender, moderately strict, angled, moderately to densely scabrous mainly along angles, with spike- lets in the distal Vi, longest branches 2.1-4.5 (-7) cm. Spikelets lanceolate, 5.3-9 miu long; glumes 2003] SORENG AND KEIL: GYNOMONOECISM & NEW POA 301 5 mm 5 cm 5 mm 5 mm FIG. 1. Poa diaboli. Illustration oF Keil 26474, holotype US; a) habit; b) ligule; c) spikelet (perfect-Howered); d) floret. 302 MADRONO [Vol. 50 narrowl)' lanceolate, 3-veined, lower (2.0?) 2.7?3.8 mm long, upper, (2.3?) 2.9?3.9 mm long; rachilla internodes 1.0-1.3 mm long, laterally visible, sparsel)' to densely, coarsely scabrous or infre- quently entirely smooth; florets (2-) 3-6 (-7); cal- luses webbed, webs diffuse, arising around the lem- ma base, hairs thin, sinuous reaching as far as mid- lemma, sparser and shorter laterally, or loosely con- centrated dorsally and absent laterally; lemmas pale green throughout, to anthocyanic along the margins to apex or throughout, lanceolate to narrowly lan- ceolate, (3.2?) 4.25-5.0 mm long, fairly thin, strongly keeled, apex acute to narrowly acute (in- frequently the keel vein excurrent as a tiny muero), margins very narrowly hyaline, surface sparsely to moderately-densely scabrous throughout, glabrous, veins 5, prominent, moderately to densely scabrous, lateral veins extending to within 1 mm of upper margins; paleas % to = lemmas in length, keels coarsely and closely scabrous, surface between keels narrow, scabrous. Flowers bisexual or pistil- late; anthers (1.4-) 1.75-2.6 mm long, or rudimen- tary and sterile; ovary glabrous, fully matured cary- opses unknown; lodicules 1.0 mm long, broadly lanceolate, long lobed. Chromosome number un- known. Flowering. March to April. Habitat. North facing slopes and ridgetops within 2-3 km of the Pacific coast, fire successional area of mesic chaparral, grassy coastal-scrub, coastal live oak woodland, and mesic Bishop pine forest, thin soils over an Edna Shale substrate, 120-400 m elev. Distribution. United States, California, San Luis Obispo Co., San Luis Range (also known as the Irish Hills). Paratypes. USA. California, San Luis Obispo Co., NW end of San Luis Range: Montana de Oro State Park, slopes W of Valencia Pk., ca. 230 m, 27 Apr 1998, D. Keil 26941 (OBI); N slopes of Valencia Pk., 35?I5'50-54"N X I20?52'19-24"W, 240 m to 383 m, 7 Apr 2003, R. Soreng, E. Frenze!, S. Krenn & J. Bleclia 7401 (US); Pacific Gas & Electric, Diablo Canyon power plant lands. North Ranch, 0.4 km E from coast road on Crowbar Can- yon Rd., 35.228?N X 120.8640?W, ca. 120 m, 19 Apr 2001, D. Keil & E. Frenzel 29331 (OBI); same loc, 120 m, 6 May 2001, E. Frenzel & D. Haines s.n. (OBI; DNA voucher); North Ranch, ridge above Coon Cr. Canyon, 35.243?N X 120.876?W, ca. 365 m, 19 Apr 2001, D. Keil & E. Frenzel 29336 (OBI); 0.3 km inland from coastal road along Crowbar Canyon Rd. at mouth of Crowbar Canyon, ca. 120 m, 14 Apr 2001, E. Frenzel DC- 19 (OBI); Ridge top S of Coon Cr., ca. 0.5 km E of "Last" benchmark and above road cut, ca. 400 m, 27 Mar 2001, E. Frenzel & D. Haines 81 (OBI). Conservation status. Narrowly endemic, locally frequent. The geographic range known from collec- tions may be less than 15 sq km. However, it occurs in very rugged country, and estimates of potential habitat suggest the range is somewhat larger. It was reportedly collected from the San Luis Range in Ruda Canyon {Glenn Clifton in 1993), but we have not been sucessful in locating the voucher. Presumably the species was overlooked because imtil recently the surrounding lands were mostly in large private holdings and extensively grazed. Light grazing of continues on some of the Diablo Canyon Power Plant land. Montana de Oro State Park (MOSP) land was privately owned until the late I960's and was grazed until the early 1970's. The coastal bench lands were used for military exercises for a short period prior to the establishment of tlie state park (Sally Krenn, Jim Blecha personal com- munication). Attempts to revegitate MOSP lands, and eradicate the invasive Ehrharta calycina (Veit Grass), included a controlled burn covering the area from the coastal bench to the top of Valencia Peak in 1997 (Vince Cicero personal communication). David Keil first noticed the new species in an area of recently burned chaparral vegetation in the fol- lowing year. The largest population of the new Poa yet discovered occurs on the peak area, and it seems the fire did not adversely impact this popu- lation, and may have benefited it. The new species epithet, diaboli (spelling veri- fied in Kunkel 1990), commemorates both its oc- currence on grounds of the Diablo Canyon Nuclear Power Plant, and secondly the difficulties it has caused the authors in determining its affinities and its breeding systein. BREEDING SYSTEM DETERMINATION Dicliny is widespread in the Poaceae (R.Br.) Barnh. (Connor 1979). However, gynomonoecism is considered uncommon, occurring in only eight genera and not in all species of tliose (Connor 1979). Connor characterized several variations in the pattern of sex expression among genera of gy- nomonoecious grasses, and all cases seem to be more or less spatially fixed in expression, and se- quentially regular. Pistillate flowers consistently oc- cur in a specific orientation, either within spikelets, or within pairs or sets of spikelets (Connor 1979), and, although there may be diffei'ces in timing of flowering, there are no reported changes in the ratio of one type of flower to another through the grow- ing season in gynomonoecious species. The varia- tions may be categorized as follows: I. Spikelets all alike, certain florets consistently pistillate. a) proximal floret pistillate, distal floret(s) perfect. b) proximal floret(s) perfect, distal floret(s) pistillate. II. Spikelets pistillate or perfect, these arranged in pairs or sets. [Vol. 50 2003] SORENG AND KEIL: GYNOMONOECISM & NEW POA 303 15 sq km. However, it occurs ?)', and estimates of potential Ige is somewliat larger. It was Tom the San Luis Range in Clifton in 1993), but we liave locating the voucher ;cies was overlooked because 3unding lands were mostly in and extensively grazed. Light in some of the Diablo Canyon /Tontana de Oro State Park ?ivately owned until the late d until the early 1970's. The 3re used for military exercises Dr to the establishment of the m, Jim Blecha personal com- s to revegitate MOSP lands, jsive Ehrharta calycina (Veit itrolled burn covering the area h to the top of Valencia Peak ro personal communication). 3d the new species in an area aparral vegetation in the fol- ?st population of the new Poa ?s on the peak area, and it t adversely impact this popu- benefited it. pithet, diaboli (spelling veri- , commemorates both its oc- jf the Diablo Canyon Nuclear ;ondly the difficulties it has determining its affinities and STEM DETERMINATION read in the Poaceae (R.Br.) '). However, gynomonoecism non, occurring in only eight 11 species of those (Connor cterlzed several variations in rression among genera of gy- s, and all cases seem to be ' fixed in expression, and se- tillate flowers consistently oc- tation, either within spikelets, s of spikelets (Connor 1979), nay be diff?r?es in timing of 3 reported changes in the ratio to another through the grow- anoecious species. The varia- ized as follows: ke, certain florets consistently jret pistillate, distal floret(s) iret(s) perfect, distal floret(s) ate or perfect, these arranged a) sessile spikelet perfect, pedicellate spike- let pistillate. b) sessile spikelet pistillate, pedicellate spikelet perfect. c) lateral spikelets pistillate, terminal spikelet perfect. Most of Connor's reports of gynomonoecism come from species or genera belonging to grass tribes or subfamilies in which the number of florets per spikelet is fixed at either one or two (subfam. Ehrhartoideae Link trib. Olyreae Kunth ex Spenn., subfam. Panicoideae Link, respectively), and where paired spikelets are common. Connor (1979) re- ports gynomonoecism in species with spikelets with indeterminate numbers of florets only from Cento- theca Desv. (subfam. Panicoideae trib. Centothe- ceae Ridl.), A^i/zi/'oa Torrey (subfam. Chloridoideae Kunth. ex Beilschm.), and Poa (subfam. Pooideae Benth.). In each of these genera the syndrome is different. Munroa fits categories la and lie (above), and Centotheca and Poa fit category lb. In Cento- theca only the proximal floret is perfect and those distal to it are pistillate. In Poa, generally only the distal most floret is pistillate, and all more proximal florets are perfect. We are not aware of any reports of changes in the frequency of pistillate flowers through time in gynomonoecious grasses. The breeding system of Poa diaboli falls on the borderline between gynomonoecism and gynodioe- cism. To characterize the breeding system of the new species, individual specimens (pre-2003) were checked for presence or absence of perfect or pis- tillate flowers (i.e., those with developed pistils and rudimentary anthers; there seem to be no staminate flowers). For this analysis individual specimens were considered to be samples with one or more synflorescences attached to a shared base, or single flowering shoots that were pressed or mounted on the same sheet and appear to be from the same individual: The individual specimens do not nec- essarily represent different plants, and more precise study of individuals through a season and through successive years is needed in the future. We assume that male sterility, as indicated by the occurrence of rudimentary anthers, is under genetic control, that anthers are not simply, aborted due to environ- mental factors (i.e., disease, moisture or nutrient stress, temperature shock, etc.; though possibly triggered by these), nor due to genetic abnormali- ties. The occurrence of rudimentary anthers is reg- ular enough within and among specimens that this seems to be a warranted assumption. Moreover, the pattern is consistent with that foiuid in other species of diclinous Poa (Soreng 1991, 2000). In geneti- cally imbalanced plants of Poa, if anther abortion occurs, it usually occurs in a late stage of devel- opment, and some pollen is formed, although usu- ally malformed to some degree (R.?S personal ob- servation). However, in diclinous species, as in the present case, anther development is halted at a much earlier, ]3resimiably pre-tapetal, stage. There are three completely pistillate specimens among an estimated 20 specimens with at least some perfect-flowers. These three specimens may represent truly pistillate plants, or late flowering plants that, had they developed earlier, would also have produced shoots with perfect-flowers, or plants in which pistillate shoots have become sep- arated from shoots with perfect-flowers in the pro- cess of collecting the plant. In seven of the speci- mens with perfect-flowers there are some panicles with a few pistillate spikelets among the lower spikelets of their synflorescences, and three speci- mens with early maturing mainly or wholly perfect- flowered panicles and late maturing, smaller, com- pletely pistillate-flowered panicles. In a survey of the Valencia Peak population in 2003 {Soreng et al. 7401) one synflorescence was taken off of 42 plants at least 2 m apart. The syn- florescence count was; three pistillate-flowered, five partly perfect-flowered/partly pistillate-flowered, and 34 perfect-flowered. Both the pre-2003 speci- men count and the 2003 population census results suggest that the new species is either intermediate between being gynomonoecious and gynodioe- cious, or, more likely, as no completly pistillate plants were found, the ratio of pistillate and perfect flowers is sequentially adjusted. A corresponding syndrome is present in a several other species of Poa. This includes co-occurrence within populations of: 1) many individuals that are perfect-flowered only, along with a high percentage of individuals (but averaging less than half) that have mixtures of perfect and pistillate flowers; 2) within mixed-flowered individuals, pistillate flow- ers occur in terminal florets within some spikelets, and often in all flowers of spikelets below the mixed-flowered spikelets, both arrangements in- creasing in frequency in late developing (lower) parts of synflorescences, and in later developing synflorescences; 3) some late developing, com- pletely pistillate synflorescences, in individuals with earlier developing perfect-flowered, or partly perfect-flowered synflorescences; and, 4) few (or no) individuals having only pistillate flowers, but then all synflorescences developing late in the breeding season. This syndrome is characteristic of the Poa sect. Homalopoa Dumort. "P. nervosa complex" of North America (i.e., P. arnowiae So- reng, P. cuspidata Nutt., P. nervo.sa (Plook.) Vasey s.stn, and P. tracyi Vasey; three other species in this complex are dioecious {P. sierrae TJ. Howell), subdioecious (P. rhizoniata Hitchc), or dioecious and gynodioecious (P. chambersii Soreng); see So- reng and Hatch 1983; Soreng 1991, 1998, 2000). The same syndrome has also been diagnosed in populations studied in the field and lab in a few South American species (e.g., P. fibrifera Pilg., an undescribed species from southern Peru and north- ern Chile, Peterson & Soreng 15615 [CONC, US]), 304 MADRONO [Vol. 50 and a few southeast Asian species (e.g., P. grandis Hand.-Maz., Soreiig, Peterson & Snn 5628; [KUN, PE, US] R.TS personal observation). Tliis breeding sj'stem, wliicli seems more lil? gy- nomonoecism, differs from true gj'nodioecism wiierein pistillate individuals are clearly distinct, occiu' in a more regular and higher proportion to perfect-flowered individuals (but not much more than half), and also mature more or less simulta- neously with those (at least in Poa; R.TS personal observation). Soreng and Hatch (1983), and Soreng (1991, 2000), previously identified this syndrome as "partial gynodioecy". However, because, in the several species with this syndrome, individual plants that are totally pistillate throughout a season are rare or uncommon, R.1S is inclined to agree with David G. Lloyd (personal communication 1992) that the syndrome is more like an extreme, and pre- viously unrecorded, form of gynomonoecism, than like weak gynodioecism. This form of gynomon- oecism is not easily distinguished from gynodioe- cism without careful examination of the distribution of different-sexed flowers in multiple populations in the field and common-garden studies, however, few totally pistillate plants are expected, along with a shift to pistillate flowers as the season progresses (Lloyd 1974, 1980). There seems to be a need to differentiate sub- types of gynomonoecism in Poa, so we shall call this system "sequentially adjusted gynomonoe- cism" to distinguish it from the common form of gynomonoecism in other Poa, and seemingly in other grasses, in wliich sex expression is fixed in spatial orientation, and is not known to change in frequency through time. Outside Poaceae, Aceva- do-Rodn'guez [2003] identified a syndrome in Tal- isia (Sapindaceae) in which individuals are "se- quentially monoecious", shifting from staminate to pistillate flower production and back to staminate flowers within a season. We are unaware of any reports of sequentially adjusted sex expression in gynmonoecious plants. Here we take up the terms "sequentially adjusted" to denote the temporal shift in sex expression through the growing season. In other gyiiomonoecious Poa, sex expression is fixed in frequency and time. In P. annna L. and other species of P. sect. Ochlopoa Asch. &. Graebn., and many South and Central American species, and some Himalayan species, pistillate flowers are con- sistently and only produced in the terminal floret, or less often also in the subterminal floret, within spikelets (Hackel 1904; Nannfeldt 1938; Chrtek and .lir?sek 1962; Connor 1979; Anton and Connor 1995). The latter syndrome is common, and may be distinguished as fixed-sex-expression gynomon- oecism. These species are not known to produce any pistillate spikelets, mostly pistillate synflores- cences, or any totally, perfect-flowered individuals. Moreover, there is no known change in frequency of pistillate florets through time, within or between plants. In their paper on "Floral biology and reproduc- tion in Poa" Anton and Connor (1995) described the many variations in Poa breeding systems known to them. They noted Soreng's application of the term "partially gynodioecious" to several North American species, but indicated their unfa- miliarity with this variation, also suggesting that it might represent an exaggerated form of gynomon- oecism. They comment on the uniformity of the syndrome of gynomonoecism in many South and Central American species of Poa, and in species of the European (now cosmopolitan) P. sect. Ochlo- poa. In these species, the lowermost 1 or 2 florets of 2 or 3 (-4) flowered spikelets are perfect, and the upper floret is pistillate. This also occurs in sev- eral Himalayan species of Poa (e.g., P. sikkimensis (Stapf) Bor; R.TS personal observation from herbar- ium specimens and Soreng, Peterson, & Sun 5676 [KUN, PE, US]). Seed is normally set in both types of flowers. They report no evidence of variation in flowering among most gynomonoecious Poa (ex- cept for P. supina Schrad. of Europe, in which the pistillate florets open one day before the perfect ones within the same spikelets). They indicate that in a few gynomonoecious species all flowers may be pistillate or perfect in some plants or synflores- cences (e.g., P. glomerifera Hack., P. Iwrridida Pilg.). The latter two species, like P. fiberifera, P. grandis, and others mentioned above, may also ex- hibit sequentially adjusted gynomonoecism (R.IS personal observation from study of herbarium spec- imens; also observed by Maria Negritto personal communication). RELATIONSHIPS AND IDENTIFICATION OF THE NEW SPECIES Although P. dial?oli seems to share sequentially adjusted gynomonoecism with species of P. sect. Homalopoa s.lat., the North American species of that group occur in forested habitats, and have much broader leaves and more loosely-tufted, erect habits, and lack hairs on the adaxial surface of in- novation blades. Also, if those have a web on the callus, as most species of P. sect. Homalopoa and other Poa do, they have a single, tight, dorsal tuft of hairs. If the new species were gynodioecious, this would support a close relationship to species of Poa sect. Madropoa Soreng, especially those of the rhizomatous subsection Madropoa Soreng. If it exhibits sequentially adjusted gynomonoecism, as we think, this species might represent an early tran- sitional element between the P. nervosa complex and P. sect. Madropoa. In comparison with species of P. sect. Madro- poa, P. diaboli is unusual in the possession, in some specimens, of thin, flat culm leaves, and fold- ed leaves that lack long, hooked or sinuous prickle hairs on and between the veins of upper leaf-blade surfaces, even on innovation shoots. In these fea- tures, it approaches P. leibergii Scribn. (of mossy. [Vol. 50 Floral biology and reproduc- md Connor (1995) described in Poa breeding systems noted Soreng's application of gynodioecious" to several ies, but indicated their unfa- ?iation, also suggesting that it :aggerated form of gynomon- ?nt on the uniformity of the inoecism in many South and cies of Poa, and in species of osmopolitan) P. sect. Ochlo- , the lowermost 1 or 2 florets red spikelets are perfect, and illate. This also occurs in sev- ;s of Poa (e.g., P. sikkiinensis onal observation from herbar- oreng, Peterson, & Sun 5676 d is normally set in both types )rt no evidence of variation in St gynomonoecious Poa (ex- hrad. of Europe, in which the 1 one day before the perfect spikelets). They indicate that cious species all flowers may ;t in some plants or synflores- nerifera Hack., P. horridula species, like P. fiberifera, P. lentioned above, may also ex- Ijusted gynomonoecism (R.TS from study of herbarium spec- ;1 by Maria Negritto personal ND IDENTIFICATION OF THE Ew SPECIES >li seems to share sequentially cism with species of P. sect, e North American species of forested habitats, and have and more loosely-tufted, erect 3 on the adaxial surface of in- o, if those have a web on the es of P. sect. Homalopoa and lave a single, tight, dorsal tuft species were gynodioecious, \ close relationship to species oa Soreng, especially those of section Madropoa Soreng. If it adjusted gynomonoecism, as 5 might represent an early tran- iveen the P. nervosa complex oa. th species of P. sect. Madro- unusual in the possession, in ;hin, flat culm leaves, and fold- Dng, hooked or sinuous prickle n the veins of upper leaf-blade inovation shoots. In these fea- P. leibergii Scribn. (of mossy. 2003] SORENG AND KEIL: GYNOMONOECISM & NEW POA 305 shaded cliffs and open slopes where snow pockets remain into early spring, of the Columbia Plateaus; gynodioecious) and P. slehbinsii Soreng (of subal- pine wet meadows, of the high Sierra Nevada; gy- nodioecious), both of P. sect. Madropoa subsect. Epiles Soreng. However, P. diaboli more often does have hairs on and between the veins of adaxial blades surfaces of its innovation shoots, like most other Madropoa species. Moi'eover, it has the ad- ditional features of a rhizomatous and/or stolonif- erous habit, and a diffusely cobwebby callus, typ- ical of species P. sect. Madropoa subsect. Madro- poa Soreng The hairs of the callus in Poa diaboli are long, soft, and sinuous, as in most Poa, but are more or less, diffusely distributed around the sides and back of the callus in many florets. The form and insertion of the callus hairs most closely resemble those found in P. confinis Vasey (of coastal dunes and adjacent sandy forests, from southernmost Alaska south to Pt. Reyes in Marin Co., California; gyno- dioecious to dioecious), P. douglasii Nees (coastal dunes, from Mendocino Co. south to Pt. Sur, Mon- terey Co., with isolated populations at Vandenberg Air Force Base, south base [a new record, D. Keil 24742], and on San Miguel and Santa Rosa Islands of Santa Barbara Co.; dioecious), and P. piperi Hitchc. (of forest openings on serpentine, endemic to the Klamath-Siskiyou region of southwest Oregon and northwest California; dioecious), all of P. subsect. Madropoa..- Poa confinis is gynodioe- cious to dioecious (n = 90, sex ratio 1.4:1, pistil- late-flowered individuals to staminate-flowered, perfect-flowered, and mixed perfect and pistillate- flowered individuals). In our estimation, despite some differences in morphology and breeding system, Poa diaboli be- longs to P. subsect. Madropoa where it approaches P. piperi, but is most similar to P. confinis. In most characteristics examined the new species is larger than P. confinis, but range of variation measure- ments often overlap. However, among other differ- ences noted above, the lemmas of P. confinis ai'e usually at least sparsely short hairy on the keel, whereas the new species is devoid of hairs (other than fine hooks or prickle hairs) on the lemma. In the treatment of Poa (Soreng 1994) in Tlie Jepson Manual: Higher Plants of California the new spe- cies keys out to a choice between P. Icelloggii and P. piperi. Different specimens key to one or the other but fit neither species description. Lead num- ber nine may be modified as follows: 9:i. 9a'. 9b. Sheaths of upper culm leaves open most of their length; innovation blades smooth or mainly sca- brous over the veins adaxially; synflorescences 10-20 cm long; flowers all perfect . . Poa ketloggii Sheaths of upper culm leaves closed over 'A of their length or more; innovation blades often hairy adaxially; synflorescences 4-10 cm long; flowers all perfect or some unisexual. Leaves thin, soft, mostly 0.8-1.5 mm wide and folded, or up to 2.5 mm wide and Hat, those of the culm gradually reduced in length upward, fre- quently flat; lemmas up to 5 mm long; plants per- fect-flowered or some (rarely all) flowers pistil- late Poa diaboli 9b'. Leaves thicker, somewhat firmer, mostly 1.5-3 mm wide, those of the culm sharply reduced in length upward, never flat; lemmas mostly 5-7 mm long; plants nearly all pistillate or slaminate- flowered Poa piperi The morphology and transitional breeding sys- tem suggest Poa diaboli is either a geologically old element in P. subsect. Madropoa, or that it is an isolated population, likely derived from P. confiius, or its progenitor, which has reverted to a more per- fect-flowered breeding system in the face of density dependent pollination factors. Few native species of Poa are known from the San' Luis Range or the adjacent and more inland Santa Lucia Range (P. howellii Vasey & Scribn., P. secunda .1. Presl subsp. secunda [P. scabrella (Thurb.) Vasey form]). The new species is readily distinguished l?om P. howellii, which occurs in the Santa Lucia Range, and in Price Canyon in the San Luis Range, by its strongly perennial habit, absence of soft hairs on the body of the lemma, diffuse cal- lus web, and its longer anthers. From P. secunda s.kit., common and widespread western North America, and sympatric with the new species (DJK personal observation; also recorded from the south end of the San Luis Range, Dudley s.n., SU), it is distinguished by the rhizoinatous habit, presence of strongly keeled lemmas with a narrowly hyaline margin, more closed culm sheaths, and presence of a well developed web on the callus, and from P. secunda subsp. secunda (the only subspecies in the region) by the lack of any soft or crisp hairs on the lemma stirface. Chloroplast DNA (cpDNA) restriction site data (Lynn Gillespie personal communication) has con- firmed the postulated relationship of P. diaboli with members of a clade including species of P. sects. Homalopoa and Madropoa, and refuted member- ship in many other sections including P. sect. Syl- vestres Soreng (including P. kelloggii, and P. rnar- cida Hitchc, also new cpDNA data) and P. sect. Secundae Soreng (including P. secunda) (see So- reng 1990, for a general cpDNA phylogenetic hy- pothesis for Poa). Of introduced species reported from San Luis Obispo Co., only the annuals Poa annua L., and P. infirma Kunth (RIS personal observation 2003), are known to occur in the area, but these have little in common with the new species. It is reasonable to postulate that the isolated geo- graphic occurrence of Poa diaboli in the south cen- tral California coastal hills represents a relictual Pleistocene distribution for species of P. sect. Mad- ropoa, as the nearest station for its closest relative, P. confinis, is over 325 km to the north. Other sec- 306 MADRONO [Vol. 50 tional members (other than P. douglasii) are dis- tributed much further north or well inland. ACKNOWLEDGMENTS We thank David Lloyd for discussions leading to the re-naming of "partial gynodioecy" a,s sequentially adjust- ed gynomonoecism, Michael Curto for identifying Keil's 1998 collections of the new species as odd and perhap.s out of range Poa piperi, and for his review, Erik Frenzel for providing some new collections and sending live plants for DNA analysis, Lynn Gillespie for providing new cpDNA data, Nancy Soreng for the illustration, Dan Nicolson for assistance with the Latin diagnosis, Paul Pe- terson, Henry Connor, and anonymous reviewers for help- ful comments on the manuscript. Sally Krenn and .lim Blecha provided background information on the history of the locals in which the species occurs, and Vince Cicero information on the 1997 burn on Valencia Peak. LITERATURE CrrED AcEVEDO-RoDR?QUEZ, P. 2003. Flora Neotropica Mono- graph 87. Melicocceae (Sapindaceae): Melicoccus and Talisia. 87:1-169. ANTON, A. M. AND H. E. CONNOR. 1995. Floral biology and reproduction in Poa (Poeae: Gramineae). Austra- lian .tournai of Botany 43:577-599. CHRTEK, J. AND V. .IiR?SEK. 1962. Contribution to the sys- tematics of species of the Poa L. genus, section Och- lopoa (A. & Gr) V. .Tiras. Preslia 34:40-68. CONNOR, H. E. 1979. Breeding systems in the grasses: a survey. New Zealand .Tournai of Botany 17:547-574. I-lACKEL, E. 1904. Zur biologie der Poa aiiiuia L. Oester- reichisches Botanische Zeitschrift 54:289-291. KUNKEL, G. 1990. Geography through botany: a dictionary of plant names with a geographical meaning. SPB Academic Publishing, The Hague, The Netherlands. LLOYD, D. G. 1974. Theoretical sex ratios of dioecious and gynodioecious angiosperms. Heredity 32:1 1-34. . 1980. Sexual strategies in plants. TIT. A quanti- tative method for describing the gender of plants. New Zealand .tournai of Botany 18:103-108. NANNFELDT, .T. A. 1938. Poa iiiacroccatia Nannf n. sp. and P. riviilaris Marie and Trabut, two more tetra- ploids of sect. Oclilopoa A. & G., and some addi- tional notes on Oclilopoa. Svensk Botanisk Tidskrift 32:295-321. SoRENG, R. .T. 1990. Chloroplast-DNA phylogenetics and biogeography in a reticulating group: study in Poa (Poaceae). American .Tournai of Botany 77:1383- 1400. . 1991. Systematics of the "Epiles" group of Poa (Poaceae). Systematic Botany 16:507-528. . 1994. Poa. Pp. 1284-1291 /;; .T. C. Tdickman (ed.). The .Tepson manual: higher plants of California. University of California Press, Berkeley, CA. . 1998. An infrageneric classification for Poa in North America, and other notes on sections, species, and subspecies, of Poa, Piiccinellia, and Di.^saiUhel- iiiiii (Poaceae). Novon 8:187-202. . 2000. Apomixis and amphimixis comparative biogeography: a study in Poa (Poaceae). Pp. 294-306 in S. W. L. Jacobs, and J. Everett (eds.). Grasses: systematics and evolution. CSIRO, Melbourne, Aus- tralia. AND S. L. HATCH. 1983. A comparison of Poa tracyi and Poa occidentalis (Poaceae: Poeae). Sida 10:123-141.