Aust. J. Agric. Res., 1996, 47, 1131-42 Characterisation of the Kangaroo Valley ecotype of perennial ryegrass (Lolium perenne) M. J. BlumenthalA, K. ~ r a k a s h A. ~ , eon forte B, P. J. ~ u n n i n ~ h aBm , and H. I. Nicol A NSW Agriculture, Pasture Research Unit, PO Box 63. Berry, NSW 2535. Victorian Department of Agriculture, Pastoral and Veterinary Institute, Private Bag 105, Hamilton. Vic. 3300. NSW Agriculture, Agricultural Research and Veterinary Centre, Forest Rd. Orange, NSW 2800. Abstract A breeding program commenced in 1992 to produce perennial ryegrass (Lolium perenne) cultivars based on the Kangaroo Valley perennial ryegrass (KVPR) ecotype with improved winter and late season growth, disease resistance, and persistence. Perennial ryegrass plants (9000) were collected from 45 sites within the Kangaroo Valley and Shoalhaven flood plain in August 1992. Using principal component analysis, site differences were greatest for N percentage, Mg and Na concentration, and A1 saturation. Collections were space planted along with standard cultivars (Banks, Ellett, Embassy, Grasslands Lincoln, and Vedette) at Berry, New South Wales (34'48's). and Timboon, Victoria (38'32'S), with 50 treatments and 10 replicates. Plants were visually scored for seasonal yield. rust incidence. greenness, leafiness, persistence, habit, and tiller density at both sites. At Berry, leaf angle. leaf width, and heading date were determined. Despite the differences in moisture, temperature, and growth indices between the 2 test sites, seasonal yield scores (mean of 100 plants) at Berry and Timboon were highly genetically correlated (r = 0.79-0.99). Standard cultivars were higher yielding with higher tiller density but with a greater rust incidence than collections (P < 0.01). Collections originating from the Shoalhaven flood plain were higher yielding with greater tiller density than plants from the Kangaroo Valley ( P < 0.01), although differences were not as great as site of origin differences would suggest. Selections have been made for polycross half-sib formation. and half-sib evaluation will take place a t a number of sites to determine the average general combining ability of parent selections for synthetic cultivar production. Additional keywords: genetic correlation. principal component analysis, site of origin. Introduction Perennial ryegrass (Lolium perenne) is the predominant grass species used in the high rainfall dairying districts of south-eastern Australia (Bartsch and Mason 1991). The Kangaroo Valley (KVPR) cultivar is an ecotype of perennial ryegrass that has evolved over 140 years of natural selection in the Kangaroo Valley and adjacent coastal flood plain from early introduction of seed from the United Kingdom (Strang 1961; Shah et al. 1990). KVPR has a number of characteristics that make it useful in plant breeding: (i) it is amongst the earliest flowering perennial ryegrasses in the world and, associated with this, is one of the most winter-active (Hill 1985; Kemp 1988); 11. J. Blumenthal et al. (ii) it is the most persistent perennial ryegrass in coastal and tableland NSW (T. Launders, unpubl. data); and (iii) it tolerates a wide range of grazing managements (Fulkerson et al. 1994). Diversity for maturity date and plant habit within the ecotype suggest that selection can be made for a range of agronomic characters. Plants grown from seed collected from seed production paddocks had a range of 23 days in the date of spike emergence, with prostrate types tending to be later flowering (Shah et al. 1990). Two cultivars (Roper and Boomer; Valley Seeds Ltd) have been developed on the basis of a small range of material from within the ecotype (Cunningham et al. 1994). Greater gains may be made from a wider collection of plant material and collection of material from old paddocks (not sown to ryegrass for at least 40 years) that have not been harvested for seed. As a first step in a breeding program, a collection of 9000 perennial ryegrass plants from 45 individual paddocks was made. This paper provides a characterisation of the collection grown at 2 sites for use in a plant breeding program and relates plant characters to site of collection. Materials and methods Collections and collection sites Nine thousand plants from 45 individual paddocks from the Kangaroo Valley and adjacent Shoalhaven flood plain were collected in August 1992. The 45 collection sites were confined within a narrow geographical area between latitudes 34'38' and 34'56's and longitudes 150°25' and 150°45'E. Many of the collection sites had not been sown to perennial ryegrass in over 100 years and all had not been sown in the last 30 years. Some of the collected material represented paddock populations that had not been used for seed harvest in the last 25 years and so represented genetic variability not present in the current KVPR cultivar gene pool. Records of annual rainfall (900-1300 mm), altitude, grazing history (dairy, cattle, sheep). grazing pressure (light, medium, high), density of perennial ryegrass (thin, satisfactory, dense), botanical composition of pasture, and soil characteristics [colour: texture; pH; P mg/kg: K, Ca, Al, hlg, Na, CEC (meq/100 g): EC (DS/m): total N percentage; A1 saturation: ESP percentage] were recorded for each collection site. Plants were collected using a 5-cm-diameter by 7.5-cm-deep corer and placed in similar sized plastic pots. Individual pots were hand-weeded, repotted in commercial potting mix, and fertilised fortnightly with Aquasol[R] for the 2 months prior t o planting out in the field. The collected plants and standard cultivars grown out in the same-sized pots (cvv. Banks. Ellett, Embassy, Grasslands Lincoln, and Vedette) were space-planted (50-cm spacing) at Berry (34'48's) and Timboon (38'32's) in October 1992. Both sites represent rain-fed dairy-farming districts. Berry has a higher proportion of rainfall in summer (December-February. 32 v. 15%), a higher January mean maximum temperature (25.6OC v. 21.0°C), and a lower July mean minimum temperature ( 6 . loC v. 7.4' C) than Timboon. Evaluation sites The Berry site was at the Pasture Research Unit situated within 50 km of the collection sites. The soil for both sites is described in Table 1. The Berry soil is classified as a red podzolic (Dr2.33; Northcote 1979) over parent rock of Berry shale with an A horizon depth of 20 cm on an imperfectly drained medium slope. The experimental site was under naturalised kikuyu (Pennisetum clandestinum) based pasture for a t least 30 years. The soil at the Timboon site is classified as a yellow podzolic (Dy3.41, Northcote 1979) over marine sedimentary material with an A horizon depth of 50 cm on an imperfectly drained medium slope. Prior t o the establishment of the experiment, a turnip crop was sown in 1989 and an oat crop in 1990. Before this, the site had been under perennial ryegrass-white clover pasture for 45 years. Characterisation of Kangaroo Valley perennial ryegrass Table 1. Soil parameters (0-10 cm) at Berry and Timboon Berry Timboon Cultural practice The experiment a t Berry was fertilised with Pasture 13 (%: 16.0 P, 12 K, 8 . 1 S, 14.7 Ca) a t 250 kg/ha and subsequently top-dressed with N in the form of urea (46% N) a t the rate of 50 kg N/ha on 29 January, 20 April, and 13 August 1993, and 8 February 1994, respectively. At Timboon, 300 kg/ha of super lime (6% P, 8% S) and 200 kg/ha of Pivot 900 NPK complex fertiliser (%: 16 N, 8 P, 9 K, 11 S) were applied before planting the trial. Subsequently, the same amount of Pivot 900 fertiliser was applied on 15 March and 7 December 1993, and 27 March and 7 September 1994. Weeds were controlled by hand-weeding and with application of Spray-seed (paraquatldiquat) and Roundup (glyphosate 360 g/L) with a weeding-wand or knapsack. Design The experiment for both sites was a completely randomised design with 50 treatments (45 collections and 5 standard cultivars) randomised over 10 blocks. Each treatment consisted of 10 individual plants spaced 50 cm apart. Measurements Estimates (1 = low, 9 = high) of dry matter yield were made prior to cutting when the highest yielding plants were a t the 3-leaf stage (when the third leaf on the tiller was fully expanded; Fulkerson and Slack 1994). Yield scores were taken 11 times between 1 March 1993 and 25 July 1994 a t Berry and 12 times between 21 January 1993 and 26 July 1994 a t Timboon. Mean seasonal (summer, autumn, winter, spring) yields were calculated for each site. Incidence of crown rust was scored (0 = no infection, 9 = high infection) on 4 occasions at Berry (4 May, 16 July 1993, 23 February, 26 April 1994) and 5 April 1993 a t Timboon. Tiller density (1 = few, 9 = many tillers) was determined on 11 February 1993 a t Berry and 16 September 1993 a t Timboon. Leafiness (1 = low, 9 = high) was determined a t Berry on 11 February 1993 and a t Timboon on 14 December 1993. Leaf angle and leaf width were measured on 12 November 1993 a t Berry. Growth habit was observed on 11 February 1993 and on 3 November 1993 a t Timboon. Greenness was estimated a t Berry on 22 December 1992, 11 February 1993, and 5 January 1994. The Berry trial was irrigated (15-20 mm/day) between 10 and 14 January 1994 t o alleviate the effects of severe drought and heat stress. Greenness was again observed on 20 January 1994 to note variation in recovery. Heading dates were observed twice weekly on selected plants that were left uncut during spring within each replicate a t Berry. A flowering score based on the number of plants heading in each row was carried out on 13 September 1993 a t Timboon. Statistical analysis All parameters were subjected t o ANOVA both with and without standards, 3- and 2-way contrasts were made between collections originating from the Shoalhaven flood plain and those originating from the Kangaroo Valley and the standard cultivars. Due t o herbicide M. J. Blumenthal et al. damage a t the Berry site, one replicate was discarded from the analysis; hence, there was an unequal number of replicates a t the sites. Because of this. the genotype and error variances at each site and the genotype covariance between sites were estimated by residual maximum likelihood (REAIL) using the program AI-RERIL (A. Gilmour, pers. comm. 1995). The genotype variances and covariance were used to compute the genotype correlation. Principal component analysis (PCA) was performed on all collection site characters and then on 4 geographical and 14 soil characters of the 45 collection sites. PCA was carried out on 4 morphological and 15 morphological-agronomic characters observed at Berry. PCA was also carried out on standardised estimates of 9 common characters measured a t Berry and Timboon, and on the combined estimate both with and without standard cultivars. Principal components were than correlated with each of the characters included in the PCA. Bivariate ellipses (P = 0.95) were fitted t o the PC1 and PC2 coordinates for collections from the Shoalhaven flood plain and from the Kangaroo Valley separately. Fig. 1. Distribution of collection sites on the first 2 principal components for the principal components analysis of soil and geographical data at the site of collection. Symbols (0, A) represent collection sites in the Shoalhaven flood plain and Kangaroo Valley, respectively. The ellipses represent the 95% confidence around the mean of each group of collections. Increasing N%, Mg a n d Na Results Collection sites Grazing history, grazing pressure density of perennial ryegrass, and botanical composition were poorly correlated with the first (PC1) and second (PC2) principal components. The results of PCA, using only 4 geographical and 14 soil characters at the collection sites, are diagrammatically represented in Fig. 1. PC1 and PC2 accounted for 45 and 17% of variability among the sites, respectively. Two distinct clusters formed: one comprising those sites in the Kangaroo Valley and the other comprising those collection sites on the Shoalhaven flood plain. All the sites were strongly acidic (pH = 4.6-5.1) and cation contents varied as follows: Na+, very low to very high; K+, low to high; h4g2+, very low to very high; and Ca2+, low to high (Hazelton and Murphy 1992). The N and P content in the soil varied from low to very high values. In general, the soils of sites in the Kangaroo Valley were less fertile than those in the Shoalhaven flood plain. Rainfall was higher in the Kangaroo Valley than the Shoalhaven flood plain (1329 v. 1166 mm). While long-term temperature data for Kangaroo Valley are unavailable, there is little variation in the temperatures between Jervis Bay on the coast and Nowra 25 km inland. The Shoalhaven flood plain, being nearer Characterisation of Kangaroo Valley perennial ryegrass to the coast, probably had higher humidity than Kangaroo Valley situated a further 25 km away. Growth indices at the two test sites Moisture, temperature, and growth indices for the experimental period (after Fitzpatrick and Nix 1970) are presented in Fig. 2. The moisture index a t Timboon followed a strongly seasonal pattern with moisture being non-limiting to perennial ryegrass growth in winter and spring. Moisture index at Berry followed a non-seasonal pattern with moisture non-limiting to perennial ryegrass growth in early summer 1992-93, autumn and parts of winter 1993, and winter 1994. Mean temperatures at Berry were generally higher than at Timboon over the experimental period with temperatures limiting perennial ryegrass growth a t both sites in winter but to a greater extent at Timboon. The growth index (which combines moisture, temperature, and light indices) was generally higher a t Berry than Timboon except over summer 1993-94 when moisture severely limited growth at Berry. Analysis of variance The 45 collections differed significantly (P < 0.01) among themselves and from the 5 standard cultivars for all characters a t both sites except for greenness and leaf angle at Berry. Collections originating from the flood plain and from the Kangaroo Valley differed for all characters at Berry except for rust score, leaf angle, and leaf width and for all characters at Timboon except yield score in autumn and winter 1994 and rust score. Genetic correlations Nine morphologic-agronomic characters common to both Berry and Timboon were subjected to AI-REML analysis for computations of genetic correlations (Table 2). Genetic correlations were higher when the standard cultivars were included in the analysis. The spring 1993 yield estimate was highly correlated between the 2 locations. The correlations for leafiness were very low. PCA analyses The variance accounted for by the first 2 principal components for the morphological and morphological-agronomic characters at Berry, the standardised estimates of 9 common characters measured at Berry and Timboon, and on the combined estimate both with and without standard cultivars are presented in Table 3. The variance accounted for ranged from 58 to 78% for PC1 and from 9 to 25% for PC2. Tiller density ( r = 0.52-0 - 97) and leafiness ( r = 0.48-0-97) were highly correlated with either PC1 or PC2 in all analyses (Figs 3 and 4). When standard cultivars were included in the analysis, they tended to have greater tiller density and leafiness than the collections (Figs 3b,d; 4b,d,f). In addition, in the morphological analysis of the Berry data, standards tended to have wider leaves and a more erect plant habit (Fig. 3b). In turn, plants originating from the Shoalhaven flood plain tended to have greater leafiness and tiller density and a more erect plant habit than the plants from Kangaroo RI. J. Blumenthal et al. Fig. 2. ( a ) Moisture, ( b ) temperature, and (c) growth indices at Berry (-) and Timboon (....) for the period of the experiment. 0 10 120 1992 Table 2. 30 40 50 1993 60 Week 170 80 90 100 110 1994 Genetic correlations between characters measured at Berry and Timboon Character Autumn 1993 yield Winter 1993 yield Spring 1993 yield Summer 1993-94 yield Autumn 1994 yield Winter 1994 yield Tiller density Leafiness Crown rust incidence With standards Without standards Characterisation of Kangaroo Valley perennial ryegrass Table 3. Percentage variability accounted for by principal components 1 and 2 for the five separate principal components analyses Analyses Berry morphological Berry morphological-agronomic Berry standardised Timboon standardised Combined standardised With standards PC1 PC2 72 63 78 69 69 21 9 10 16 11 Without standards PC1 PC2 67 58 63 61 61 25 11 19 20 13 Valley (P < 0.01; Figs 3a,c; 4a,c,e). In the Berry morphological-agronomic analysis, PC1 was highly correlated (r = 0.87-0 91) with persistency (Fig. 3 b, d) . Persistency (Fig. 3 b, d) (r = 0 73-0.84), leafiness (r = 0.45-0 83), tiller density (r = 0.44-0.82), and plant habit ( r = 0.51-0.72) were all highly correlated with yield estimates. Yield ( r = 0.45-0.97) and rust (r = 0 - 71-87) incidence were highly correlated with PC1 in the Berry morphological-agronomic analysis and the analyses using the standardised estimates common to both Berry and Timboon. The standard cultivars tended to have higher yield and greater rust incidence than the collections (P < 0.01; Figs 3d, 4d-f). The yield performance of 2 of the collections was similar t o that of 4 of the 5 standard cultivars (Fig. 4f). In turn, plants from the Shoalhaven flood plain had higher yield than plants from Kangaroo Valley (P < 0.01). Using the standardised Timboon (Fig. 4c,d) and the combined data with standards (Fig. 4f), PC2 was highly correlated ( r = 0.56-0.87) with autumn 1994 yield. Using the standardised combined data without standards, PC2 was highly correlated (r=0.56) with winter 1994 yield (Fig. 4e). Correlations between plant characters and collection site In all data sets, soil P and N percentage at the collection site had significant correlations with mean seasonal yield scores (P < 0 e01). The correlation ( r ) of yield score with soil P content ranged from 0.439 to 0.513 (Berry), 0.433 to 0.584 (Timboon), and from 0.483 to 0.583 (combined); and for total N ranged from 0-489 to 0.608 (Berry), from 0.404 to 0.495 (Timboon), and from 0.487 to 0.491 (combined). Magnesium percentage was correlated with tiller density: r = 0.540 ( P < 0.01; Berry), 0 -555 ( P < 0.01; Timboon), and 0.501 (P < 0.01; combined); and with leafiness: r = 0.490 (P < 0.01; Berry) and 0.431 (P < 0.01; combined). At Berry, total soil nitrogen at the collection site was correlated with greenness (0.485; P < 0 . Ol), tiller density (0.584; P < 0. Ol), and leafiness (0.532; P < 0.01). Discussion Three main points emerge from this work. (i) Despite the differences in moisture, temperature, and growth indices between the 2 sites, the performance of the collections and plants from the standard cultivars was highly correlated between Berry and Timboon. (ii) Standards tended to be higher yielding and have higher tiller density, but greater rust incidence, than the collections. (iii) Collections from the Shoalhaven flood plain tended to be higher yielding with greater tiller density than plants from the Kangaroo Valley, although differences were not as great as differences in site of origin would suggest. M. J. Blumenthal et al. Increasing leafiness and tiller density Increasing yield, rust and persistency Fig. 3. Distribution of collections and standard cultivars (Ba, Banks; El, Ellett: Em, Embassy; Ln, Grasslands Lincoln: and Ve, Vedette) on the first 2 principal components for the principal components analysis of morphological data only ( a ) without standards and ( b ) with standards: and of morphologic-agronomic data (c) without standards and (d) with standards a t Berry. Symbols (0, A) represent collection sites in the Shoalhaven flood plain and Kangaroo Valley, respectively. The ellipses represent the 95% confidence around the mean of each group of collections. Genetic correlation between test sites Despite the differences in moisture, temperature, and growth indices between the 2 sites, the performance of the collections and standard plants was highly correlated between Berry and Timboon. At the start of the experiment, we thought that the non-seasonal Berry environment and the more I\Iediterranean Timboon environment would impose moisture and temperature stresses at different times of the year and we predicted an environment xgenotype interaction, with collections performing differently at each site. This assumes sufficient variability between collections in adaptive plant characters. Certainly, significant differences did occur in plant characters, but there was no interaction with test environment. This is surprising as moisture stress did occur at different times in the year at . . . . . " * " ... - Characterisation of Kangaroo Valley perennial ryegrass Increasing yeild, tiller density, and rust Fig. 4. Distribution of collections and standard cultivars (Ba, Banks; El. Ellett: Em, Embassy; Ln, Grasslands Lincoln: and Ve, Vedette) on the first 2 principal components for the principal components analysis of the 9 characters in common at the Berry and Timboon sites: ( a ) without standards and ( b ) with standards at Berry, (c) without standards and (d) with standards at Timboon, and ( e ) without standards and ( f ) with standards for the combined data. Symbols (U, A) represent collection sites in the Shoalhaven flood plain and Kangaroo Valley, respectively. The ellipses represent the 95% confidence around the mean of each group of collections. M. J. Blumenthal et al. a t Timboon, moisture stress occurred only in summer. The large within-collection variability may mean that each collection has enormous plasticity, with mean collection performance being similar in contrasting environments. Clearly, there is scope for the development of a highly robust, widely adapted cultivar based on these collections. In a study of perennial ryegrass collected from a more diverse range of environments than in this study, Wedderburn et al. (1989) were not able to differentiate distinct ecotypes until the collections were stressed by either low fertility or moisture deficit (Wedderburn et al. 1990). Imposition of stress by growth in soils of contrasting fertility and the use of rain-out shelters may help to differentiate the performance of individual collections. Performance of standard cultivars The standard cultivars were generally higher yielding than the collections but were more rust susceptible. They were also more erect and leafier, had wider leaves, and were more densely tillered and more persistent over the 2 years of the experiment. All the standards were derived from New Zealand ecotypes or from hybridisation between selections within New Zealand ecotypes and were all later flowering than the collections. Late flowering lines often yield less in winter (Widdup and Ryan 1993). Despite this, they were higher yielding in the first autumn and winter than the collections. Ellett and Grasslands Lincoln (essentially a Ruanui type with low endophyte) were also used as standards in a New Zealand North Island hill country collection (Wedderburn et al. 1989) and an extensive South Island collection (Widdup and Ryan 1992). Compared with plants collected on the North Island, the standards were also more erect, had fewer tillers, and had larger and wider leaves but were only slightly later flowering. In contrast, when the standards were compared to the plants collected on the South Island, they were still more erect with longer and wider leaves, had fewer tillers, and had a similar flowering date. In our study, from the first spring onwards, only one of the standards (Vedette) had yield scores higher than the best of the collections. Interestingly, Vedette has not performed as well as other cultivars, including KVPR, when grown in swards throughout coastal NSW (T. Launders, unpubl. data). Crown rust susceptibility was correlated with yield ( r = 0 -42-0 .go), tiller density (0 6 2 ) , leafiness (O.72), plant habit (O.52), and maturity (0.51). This contrasts with findings of Humphreys (1989) and Balfourier and Charmet (1991) where rust susceptibility was negatively correlated with heading date. These results do not support Balfourier and Charmet's (1991) suggestion that a prostrate growth habit favours rust infection; the opposite is the case in this study with more erect plants being more susceptible to rust. Other field studies have shown that the more erect standards (including Ellett and Ruanui) were less rust susceptible than the more prostrate collections from the South Island of New Zealand (Widdup and Ryan 1992). The epidemiology of crown rust infection is complex and the genotype of perennial ryegrass would certainly affect the development of crown rust regardless of plant habit. We suggest that the KVPR ecotype has developed resistance to rust strains peculiar t o eastern Australia. Characterisation of Kangaroo Valley perennial ryegrass Plant characters and site of origin within the KVPR ecotype Plants collected from the Shoalhaven flood plain were higher yielding, later flowering, and more erect, and had greater tiller density, than plants collected from the Kangaroo Valley proper. Soil fertility, particularly total N and Mg, and Na cations were the site of origin characters differing most between collection sites, with flood plain sites being generally higher in N and exchangeable cations. Yield was correlated with soil P and N at the site of origin. This is in contrast to work in New Zealand where plants collected from low P sites were more vigorous than plants from high P sites (Forde and Suckling 1980). Tiller density and greenness were also correlated with Mg level at the site of origin. Plant characters were independent of other site of origin characteristics, including past management and climatic factors. Whilst site of origin characters were strongly differentiated between collection sites on the Shoalhaven flood plain and sites in the Kangaroo Valley, the bivariate ellipses suggest that plants collected from the 2 locations formed a single normally distributed population even though the large number of plants sampled meant that plants from the 2 locations were statistically different. Future breeding Selections have been made on the basis of individual plant performance, with emphasis on the 2 collections that performed as well as the standards, for half-sib formation. Half-sib evaluation a t a number of sites will be used to determine the average general combining ability of parent selections for synthetic cultivar production. The superior performance of the standards and a few of the collections suggests that hybridisation with foreign germplasm not previously introduced into the Kangaroo Valley collection would be of merit, particularly in combining high winter and late season growth. The performance of European tetraploids in Europe (Humphreys 1991) and in New Zealand (Widdup and Ryan 1993) suggests that induced polyploidy within adapted diploid ecotypes and cultivars of perennial ryegrass may lead to gains in both forage yield and quality. Tetraploidy appears t o be particularly beneficial for the combination of both winter and late season growth (Widdup and Ryan 1993). The Kangaroo Valley collection is currently being assessed for presence of endophyte; levels of peramine, ergovaline, and lolitrem; Mg concentration; dry matter digestibility; and soluble carbohydrate levels, so that issues of feed quality can also be addressed within the plant breeding program. Acknowledgments This work has been funded by the Australian Dairy Research Corporation and the Kangaroo Valley Perennial Ryegrass Seedgrowers Co-operative Ltd. The local knowledge of Brian Morrison and the historical records freely provided by farmers made the collection possible. Geoff Rowe, Malcolm Anderson, Bruce Ardler, and Brian Morrison assisted in the plant collection. References Balfourier, I?., and Charmet, G. (1991). Relationships between agronomic characters and ecogeographical factors in a collection of French perennial ryegrass populations. Agronomic 11, 645-7. M. J. Blumenthal et al. Bartsch, B., and Mason, W. (1991). 'Feedbase 2000'. (Dairy Research and Development Corporation: Glen Iris, Vic.) Cunningham, P. J., Blumenthal, M. J., Anderson, M. W., Prakash, K. S., and Leonforte, A. (1994). Perennial ryegrass improvement in Australia. New Zealand Journal of Agricultural Research 37, 295-310. Fitzpatrick, E. A., and Nix, H. A. (1970). The climatic factor in Australian grassland ecology. In 'Australian Grassland'. (Ed. R. M. Moore.) pp. 3-26. (Australian National University Press: Canberra.) Forde, M. B., and Suckling, F. E. T. (1980). Genetic resources in high-rainfall hill pastures in New Zealand. 11. Description of the ryegrass collection. New Zealand Journal of Agricultural Research 23, 179-89. Fulkerson, W. J., and Slack, K. (1994). Leaf number as a criterion for determining defoliation time in Lolium perenne. 1. Effect of water-soluble carbohydrates and senescence. Grass and Forage Science 49, 373-7. Fulkerson, W. J., Slack, K., and Lowe, K. F. (1994). Variation in response of Lolium genotypes to defoliation. Australian Journal of Agricultural Research 45, 1309-17. Hazelton, P. A., and Murphy. B. W. (1992). 'What do all the Numbers Mean?-A Guide for the Interpretation of Soil Test Results.' (Department of Conservation and Land Management: Sydney.) Hill, B. D. (1985). Growth patterns of perennial ryegrass varieties. In 'Proceedings of 3rd Australian Agronomy Conference'. Hobart, p. 192. Humphreys, M. 0. (1989). Assessment of perennial ryegrass (Lolium perenne L.) for breeding. 11. Components of winter hardiness. Euphytica 38, 75-84. Humphreys, M. 0. (1991). The value of polyploidy in breeding hybrid grasses. In 'Proceedings of 17th Meeting of the Fodder Crops Section of Eucarpia'. Algbero, Italy, pp. 27-44. Kemp, D. R. (1988). The effects of flowering and leaf area on sward growth in winter of temperate pasture grasses. Australian Journal of Agricultural Research 39, 597-604. Northcote, K. H. (1979). 'A Factual Key for the Recognition of Australian Soils'. 4th Edn. (Rellim Technical Publications: Glenside, S. Aust.) Shah, G. H., Pearson, C. J., and Read, J. W. (1990). Variability in habit, flowering and seed production within the Kangaroo Valley cultivar of Lolium perenne when grown in a range of environments. Australian Journal of Agricultural Research 41, 901-9. Strang, J. (1961). Kangaroo Valley perennial ryegrass. The Agricultural Gazette NSW 72, 131-3. Wedderburn, M. E., Pengelly, W. J., Tucker, M. A., and di Menna, M. E. (1989). Description of ryegrass removed from New Zealand North Island hill country. New Zealand Journal of Agricultural Research 32, 521-9. Wedderburn, M. E., Pengelly, W. J., Tucker, M. A., and Ledgard, S. F. (1990). Response of a New Zealand North Island hill perennial ryegrass collection to nitrogen, moisture stress, and grass grub (Costelytra zealandica) infestation. New Zealand Journal of Agricultural Research 33, 405-1 1. Widdup, K. H., and Ryan, D. L. (1992). Forage potential of wild populations of perennial ryegrass collected from southern New Zealand farms. Proceedings of New Zealand Grassland Association 54, 161-5. Widdup, K. H., and Ryan, D. L. (1993). Potential of European perennial ryegrass germplasm in southern New Zealand. In 'Proceedings 17th International Grassland Congress.' Palmerston North and Rockhampton, pp. 209-11. Manuscript received 8 December, accepted 21 May 1996