Softening of temperate annual pasture legume hard seeds on the soil surface and with shallow burial at three contrasting sites in southern Australia

Experimental sites

This experiment was conducted at the following three sites across southern Australia:

Shenton Park and Urrbrae are located close to the coast (5 km and 10 km inland respectively), whereas Ginninderra is situated 150 km inland in the Southern Tablelands of NSW/ACT. The soil in Perth was a deep sand (Kandosol), into which loam had been incorporated, the soil in Canberra was a Yellow Chromosol sandy loam, whereas the Adelaide site was a Red Chromosol with a sandy loam topsoil. Soil pH (in CaCl₂) of the top 20 cm was 6.0 in Perth, 6.2 in Adelaide and 5.5 in Canberra.

Air temperatures and rainfall records from the nearest Australian Bureau of Meteorology recording station were used to compare site differences. These corresponded to Perth metropolitan (Station 009225) for Perth, Kent Town (Station 023090) for Adelaide and Ginninderra (Station 070350) for Canberra. At each site, a temperature button, with hourly temperature recordings, was placed on the soil surface from the time of seed placement to record soil surface temperatures, whereas a second button was placed 2 cm below the surface from March to measure soil temperatures of buried seed samples. However, technical difficulties meant that only data from Perth were obtained.

Production of seeds for hard seed tests

In total, 42 annual legume cultivars were evaluated. These consisted of 15 subterranean clovers (Trifolium subterraneum), nine other clovers (Trifolium species), 11 serradellas (Ornithopus species), six annual medics (Medicago species) and biserrula (Biserrula pelecinus) (Table 1). Twenty cultivars, referred to in this study as ‘core cultivars’, were gown in common at all three sites, six were grown in both Perth and Canberra and three were grown in both Perth and Adelaide, whereas eight were grown only in Perth, four only in Canberra and one only in Adelaide (Table 1). Seeds for hard seed studies were obtained from plants grown at their respective sites during the 2016 winter–spring growing season. Common seed lots were sown at the three sites. Clover, serradella and biserrula seeds were obtained from the Department of Primary Industries and Regional Development, Western Australia, whereas annual medic seeds were obtained from the South Australian Research and Development Institute.

Scarified seeds were sown into Petri dishes moistened with filter paper and placed in a 15°C temperature cabinet for 3 days. Sowing dates in 2016 were 27 June at Perth and 29 June at Adelaide. Seeds were initially sown on 30 June at Canberra, but poor germination of some cultivars resulted in a subsequent sowing on 22 July, with a further sowing of O. compressus cv. Avila on 8 August. The late sowing time was aimed at compressing flowering time differences to allow sufficient time for late-flowering varieties to set seed before the early flowering varieties senesced. Germinants with visible radicles were transplanted into peat Jiffy pots in a glasshouse and the appropriate Rhizobium strains for each species were applied by watering on a slurry of peat inoculum (Group C for Trifolium species, Group AM for Medicago polymorpha and M. truncatula, Group AL for M. littoralis, Group S for Ornithopus species and Group Bis for B. pelecinus). Seedlings were watered daily and transplanted to the field 5 weeks after germination. In Canberra, all available seedlings from the 30 June sowing were transplanted with additional seedlings, where needed, from the 28 July sowing and from the 8 August sowing for cv. Avila.

Trial design and management

Weed-free seed beds were prepared by cultivation and application of glyphosate herbicide at label rates. Each trial had a randomised block design with three replicates. To minimise weed growth, seedlings were transplanted into 10 cm diameter holes cut into 1200 mm wide × 30 μm thick strips of white plastic mulch film (Integrated Packaging, Resevoir, Victoria, Australia) in Perth and into black woven polypropylene matting (Permathene Noweed Premium Weed Mat, Castle Hill, New South Wales, Australia) in Adelaide and Canberra. Plots consisted of 20 plants. In Perth and Canberra, they consisted of two rows of 10 plants, whereas plots in Adelaide consisted of four rows of five plants. Plant spacing in each direction within plots was 20 cm, with plots separated by 1 m. In Perth, single superphosphate with potash (6.8% P, 12.4% K and 8.3% S) was drilled in just prior to transplanting at a rate of 180 kg/ha and was also applied by hand at the same rate on 15 September. However, the sites at Canberra and Adelaide were not fertilised, because of high-fertility conditions following crop experiments. No symptoms of any nutrient deficiencies were observed at any of the sites. The sites were hand-weeded throughout the experiment, and Bifenthrin (100 mL/ha) was applied on 9 September to control redlegged earth mites (Halotydeus destructor) at Canberra.

Sites were rainfed, but the Perth trial had supplementary irrigation from 10 October until 23 November, to enable later-flowering cultivars to set seed. Monthly rainfall and temperature data for the three sites from 1 July 2016 to 30 June 2017 are shown in Fig. 1.

Fig. 1.

Monthly rainfall (mm) and temperature (°C) data for (a) Perth, (b) Adelaide, and (c) Canberra. Rainfall for the experimental period of 1 July 2016–30 June 2017 is shown as bars. Mean monthly maximum and minimum air temperatures during the same period are shown as red and blue lines, respectively. Data from the closest Australian Bureau of Meteorology stations: Perth metropolitan (Station 009225) for Perth; Kent Town (Station 023090) for Adelaide and Ginninderra (Station 070350) for Canberra.

To enable subterranean clovers to bury their burrs for normal seed development (Collins et al. 1976), the plastic was cut away from subterranean clover plots prior to flowering in Perth, whereas a 2 cm layer of weed-free sand was applied on top of the polypropylene matting in Adelaide and Canberra. The plastic strips or polypropylene matting remained in place for other species.

Flowering and senescence

Days to first flowering (DFF) was recorded as the number of days from germination to when 50% of plants in each plot had at least one open flower. Plants were checked every 3–4 days. Time to senescence was also recorded at Perth, determined as the date when no green leaves, petioles or stems remained in the plot. This allowed the reproductive period to be calculated as the difference between the number of days from first flowering to senescence.

Seed harvesting and processing

Fruits (burrs, pods or seed heads) were harvested when plants were fully senesced. In Perth, plants were harvested on 3 January, except for Cefalu arrowleaf clover and Electra purple clover, which did not fully senesce until 12 January, and Arrotas arrowleaf clover, which was too late flowering to produce seeds. Plants in Adelaide were also harvested on 3 January, whereas plants were harvested in Canberra on 31 January. Sampling was focussed on fruits closest to the crowns, because these were considered likely to be the most mature. Sufficient fruits were harvested from each plot to provide 700 seeds for testing.

Prior to seed processing, fruits were placed in an oven at 35°C overnight for easier seed extraction. Soil particles were first gently removed by sieving. The seeds of most species were readily extracted from fruits by gentle rubbing onto ribbed rubber matting. Debris was then removed by sieving to give a clean seed sample. However, whole pods of serradellas and barrel medics were used for hard seed testing, because seeds were too difficult to extract from their pods without causing significant scarification or damage. Samples of harvested material were stored at ambient room temperatures before and after seed processing.

Seed samples for hard seed testing

Seed samples from each plot were divided into seven lots of 100 seeds to allow the following seven sampling times: (i) a sample to determine initial hardseededness at field ripeness; (ii) five sampling times every 28 days over the summer–autumn period (from February to June); and (iii) a sample buried 2 cm below the surface from February to June. Serradella pod segments (each containing a single seed) were counted and allocated similarly to each sampling time. For barrel medics, 15 pods were allocated to each treatment; this was shown to be sufficient for at least 100 seeds.

Seed and pod samples for hard seed testing were placed inside 60 mm × 60 mm pockets of nylon insect screen (0.88 mm mesh) formed using a laboratory heat sealer, using the procedure of Revell et al. (1998). Pockets were arranged in strips in the same order as the seed production plots and the edges were heat sealed to prevent seed escape. The strips of pockets were then pinned flat to the soil surface in full sunlight on a patch of cleared soil. In this way seeds and pods formed a single layer with good soil contact. For the buried treatment, one seed lot was buried to 2 cm at the same time the February sample was measured, following prior exposure on the soil surface. This was aimed at simulating shallow cultivation or trampling by stock.

Germination tests

An initial germination test was conducted at the time when samples were placed in the field. Remaining samples were removed from the field for germination testing every 28 days over the summer–autumn period. Sampling dates at each site were 23 February, 23 March, 20 April, 18 May and 15 June. Particular focus was given to two sampling times. From the Chapman and Asseng (2001) analyses of false break occurrences in south-western Australia, 23 March was used as the date when seeds were considered to be most prone to germination from a false break. The final sampling date of 15 June was used to determine the proportion of residual hard seeds that remained in the seed bank as dormant seeds for germination in future years.

Germination tests were conducted according to standards of the International Seed Testing Association (ISTA 2025). Seed and pod samples were placed in Petri dishes (10 cm diameter) containing a sheet of Whatman No. 1 filter paper moistened with 3.5 mL of filtered water. Petri dishes were placed inside sealed ziplock plastic bags (325 mm × 230 mm) in the dark at a constant temperature of 15°C for 14 days. Preliminary counts were made 7 days after wetting up, with a second count after 14 days.

For the initial (freshly harvested) sample, seeds were classified into the following categories: (i) germinated (seed imbibed with an emerged radical at least 2 mm long); (ii) hard (seed not imbibed); (iii) dormant (seed imbibed with no emerged radical but firm when prodded with forceps); or (iv) dead (seed imbibed with no emerged radical but soft when prodded with forceps, often accompanied by fungal attack). However, the procedure differed for barrel medics and serradellas, whereby germinated seeds erupting from pods were first counted. Barrel medic pods were then manually opened and the remaining seeds inside were counted and classified as above. For serradellas, hard seeds were distinguished from soft seeds by prodding pod segments with forceps. It was assumed that all soft, ungerminated serradella seeds were dead, because they could not be distinguished from dormant seeds. Only hard seeds were counted for subsequent samples removed from the field, because some germination of softened seeds occurred following rainfall events over the summer–autumn period. Hard seeds remaining after 14 days of imbibition were taken as the measure of hardseededness for all genotype, time and site comparisons.

Statistical analyses

Two-way ANOVAs for percentage hard seed were conducted for each site, with cultivar and sampling time as main effects. Two-way ANOVAs were also conducted to compare days to first flowering and percentage of hard seed remaining on 23 March (to simulate the effect of a false break) and on 15 June (to estimate residual hard seeds remaining in the seed bank) for the 20 core cultivars common to the three sites, using cultivar and site as main effects. Additional entries common to Perth and Adelaide and to Perth and Canberra were also compared in this way. To compare the effects of burr burial on residual hard seeds remaining on 15 June, a three-way ANOVA across the 20 core cultivars common to the three sites was conducted, with site, burial treatment and cultivar as factors. Differences among means and the interactions among variables were assessed using least significant differences (P = 0.05). All analyses were conducted using Genstat (ver. 23.1, VSN International Ltd).

Climatic conditions

Rainfall received during 1 July−20 November, corresponding to the period of growth following transplantation, flowering and seed set, was similar at the three sites, with 475 mm at Perth, 417 mm at Adelaide and 455 mm at Canberra (Fig. 1). This was non-limiting for growth and seed set, particularly with supplementary November irrigation in Perth. Rainfall post-seed set from 1 December to 30 April differed markedly among the sites, being 111 mm at Perth, 184 mm at Adelaide and 265 mm at Canberra (Fig. 1).

Canberra was considerably cooler than Perth and Adelaide during the 2016 winter and spring growing season (July–November), reflecting its higher altitude (Fig. 1). This was most pronounced for mean minimum winter and spring temperatures at Canberra, which were 1.9°C and 6.4°C respectively, compared with 7.7°C and 10.6°C respectively, at Perth and 7.9°C and 10.6°C respectively, at Adelaide. Mean maximum temperatures at Canberra during these periods were also 3–4°C cooler than those at the other two sites (Fig. 1).

Mean summer (December 2016–February 2017) maximum temperatures following seed ripening were similar at the three sites, but the mean summer minimum temperature at Canberra (14.0°C) was ~3°C cooler than both Perth and Adelaide. Mean autumn (March–May) maximum temperatures in 2017 varied across the sites, being 25.9°C at Perth, 23.9°C at Adelaide and 20.4°C at Canberra (Fig. 1). The mean autumn minimum temperature at Canberra of 6.9°C was also considerably lower than that in Perth and Adelaide (both 13.4°C). The soil surface at Perth had a 1.6°C higher mean temperature than at 2 cm depth from 1 March to 13 June 2017 (Table 2). The surface temperature also had a higher amplitude, with a 20.1°C difference between the mean maximum and mean minimum temperatures, compared with 16.3°C at 2 cm depth (Table 2).

Flowering and seed set

Days from sowing to first flowering at Perth, Adelaide and Canberra are shown in Table 3. There was a highly significant (P < 0.001) cultivar effect for flowering times at each site, with a large range among and within species. The annual medics had the earliest flowering cultivars at each site, whereas Tammin and Leura were the earliest and latest flowering subterranean clovers respectively. Cadiz and Margurita were the earliest flowering French serradellas, whereas Serratas was the latest. Of the yellow serradellas, Yelbini was the earliest flowering cultivar at Perth and SerraMax the earliest flowering at Canberra and Adelaide, whereas Avila was the latest flowering at Perth and Canberra. Of the other species, gland clover at Perth and Frontier balansa clover at Perth and Adelaide were the earliest flowering. However, Arrotas arrowleaf clover did not flower at Perth until 7 December (37 days later than did Leura subterranean clover) and failed to set seed. It was consequently excluded from hard seed screening. Senescence dates and calculated reproductive periods at Perth are shown in Supplementary Table S1.

A two-way ANOVA for flowering time of the 20 core cultivars common to each site showed a highly significant (P < 0.001) site effect, with a mean flowering time of 105 days from sowing for both Perth and Adelaide and 116 days for Canberra. Cultivar and site × cultivar interactions were both highly significant (P < 0.001) (Table 3).

Initial hard seed levels

The hard seed level at field ripening of Cadiz French serradella was ≤0.3% at all three sites and for Serratas French serradella it was ≤6.5% at both Perth and Canberra. At Perth, all other varieties had >85% initial hard seed levels, apart from Avila yellow serradella (73.8%), whereas at Adelaide, all other varieties had >80% initial hard seed levels, apart from Margurita French serradella (33.0%) and Yellotas yellow serradella (37.7%). At Canberra, initial hard seed levels (measured in February) of all other varieties were >80%, apart from Erica French serradella (32.0%), subterranean clovers Seaton Park (66.7%), Monti (71.7%), Gosse (72.7%) and Larisa (75.3%), and Herald strand medic (78.7%).

Hard seed softening over summer–autumn on the soil surface

Graphs of hard seed softening on the soil surface over the summer–autumn period in Perth, Adelaide and Canberra are shown in Figs 2, 3, and 4 respectively.

Fig. 2.

Hard seed softening of annual legume cultivars on the soil surface over the summer–autumn period at Perth: (a) subterranean clovers (b) serradellas; (c) annual medics; and (d) other species. Values are means of three replicates. Error bars represent standard errors. Cultivar effect across all sampling times: P < 0.001; l.s.d. = 2.7; sample time effect across all cultivars: P < 0.001; l.s.d. = 1.1; cultivar × sample time interaction: P < 0.001; l.s.d. = 6.5.

Fig. 3.

Hard seed softening of annual legume cultivars on the soil surface over the summer–autumn period at Adelaide: (a) subterranean clovers (b) serradellas; (c) annual medics; and (d) other species. Values are means of three replicates. Error bars represent standard errors. Cultivar effect across all sampling times: P < 0.001; l.s.d. = 4.5; sample time effect across all cultivars: P < 0.001; l.s.d. = 2.3; cultivar × sample time interaction: P < 0.001; l.s.d. = 11.0.

Fig. 4.

Hard seed softening of annual legume cultivars on the soil surface over the summer–autumn period at Canberra: (a) subterranean clovers (b) serradellas; (c) annual medics; and (d) other species. Values are means of three replicates. Error bars represent standard errors. Cultivar effect across all sampling times: P < 0.001; l.s.d. = 4.3; sample time effect across all cultivars: P < 0.001; l.s.d. = 1.8; cultivar × sample time interaction: P < 0.001; l.s.d. = 9.7.

Seed softening of buried seeds

Comparisons of the residual hard seeds on 15 June of all cultivars at each site, following placement either on the surface or buried at 2 cm depth, are shown in Fig. 5.

Fig. 5.

Hard seed levels of annual legume seeds on 15 June following seed softening over the summer–autumn period either placed on the soil surface (blue bars) or buried at 2 cm depth (orange bars) at (a) Perth, (b) Adelaide, or (c) Canberra. Values are means of three replicates. Error bars represent standard errors. Unsown cultivars are designated by x.

The 20 core cultivars common to the three sites were analysed in more detail to determine the effects of seed burial at different sites. A preliminary ANOVA for the nine common subterranean clover cultivars showed a highly significant (P < 0.001) cultivar effect, but neither subspecies nor subspecies by burial treatment effects were significant. The subspecies were therefore grouped together in subsequent analyses. However, there was a highly significant difference between the two serradella species (P < 0.001), so consequently French and yellow serradellas were treated as separate groups in subsequent analyses. There were insufficient representatives of the three annual medic and three alternative legume species to subdivide into species groups, so single groups were used for comparisons, even though it was recognised that balansa clover, bladder clover and biserrula showed different seed-softening dynamics. Results from a three-way ANOVA (with site, burial treatment and species group as factors) across the 20 core cultivars is shown in Table 4, with species groups of subterranean clover, annual medics, French serradella, yellow serradella and other species.

Across all species groups, burial at 2 cm depth significantly (P < 0.05) reduced the mean amount of seed softening over the three sites. The site effect was highly significant (P < 0.001), with Perth showing the greatest difference between burial treatments, but neither the site × burial treatment nor the three-way site × species group × burial treatment interactions were significant (Table 4). However, there was a highly significant (P < 0.001) species group × burial treatment interaction. The overall reduction in seed softening among species groups can be attributed to the high number of subterranean clover entries (9), which had a 37.0% reduction in seed softening following shallow burial. This was particularly noticeable at Perth, with a three-fold reduction, and at Adelaide, with a 38.4% reduction. Reductions in seed softening following burial can generally be seen among the other non-core subterranean clover cultivars, particularly at Perth (Fig. 5).

However, pod burial had the opposite effect on the two core yellow serradellas, which showed a 29.2% mean increase in seed softening with shallow burial. This was most noticeable at Perth (51.1% increase) and at Adelaide (37.4% increase), whereas burial at Canberra had little effect. This difference in seed softening following pod burial was largely attributable to a 71.7% mean reduction in hardseededness ofcv. SerraMax (Fig. 5), which was particularly pronounced in Perth (89.9% reduction) and Adelaide (53.1% reduction), compared with Canberra (9.6% reduction). Significantly increased seed softening following pod burial was also observed in the other non-core yellow serradellas at Perth and Adelaide, and with cv. Avila at Canberra, apart from cv. Yellotas at each site and cv. Santorini at Canberra (Fig. 5).

There were no significant differences in the mean response to shallow burial in the core cultivar groups of French serradella, annual medics or other species across the three sites (Table 4). However, there were significant reductions in seed softening following burial at an individual cultivar level in the balansa clovers Paradana at Perth (but not Adelaide and Canberra) and Frontier at Adelaide (but not Perth), Casbah biserrula at Adelaide (but not the other two sites) and in Electra purple clover, Sothis eastern star clover and Cefalu arrowleaf clover at Perth (Fig. 5). Scimitar burr medic also had significantly less softening following burial at Perth (but not Adelaide and Canberra), whereas Herald strand medic softened more with burial at Perth, but less at Canberra (Fig. 5). Also of note was the inconsistent reaction to burial of Bartolo bladder clover across the three sites, with greater seed softening than surface-placed seeds at Adelaide, less seed softening at Canberra and negligible differences at Perth.

Cultivar differences for hardseededness

Considerable genetic variation was demonstrated among the annual legumes for both the extent of residual hard seeds carried over in the seed bank for germination in subsequent seasons and for tolerance to false breaks, owing to differences in the timing of hard seed softening over the summer–autumn period. This diversity conforms with previous studies, including Norman et al. (2002, 2006), Loi et al. (1999, 2005), Jeffery et al. (2021), Howieson et al. (2021), Nutt et al. (2021a) and Harrison et al. (2024) and indicates that considerable gains can be made by selection of annual pasture legumes for both traits to match the requirements needed for different target environments and farming systems. Across the sites, the soft seeded Cadiz and Serratas French serradellas had negligible hard seeds at seed maturity, whereas the French serradellas bred for hardseededness (Margurita and Erica) demonstrated higher initial and residual hard seed contents, similarly to the experiments of Howieson et al. (2021), Newell et al. (2023) and Harrison et al. (2024). Of the other annual legumes, subterranean clovers had the greatest amount of seed softening over the summer–autumn period, in accordance with the studies of Evans and Smith (1999), Loi et al. (2005), Norman et al. (2006), Nichols et al. (2013), Howieson et al. (2021) and Newell et al. (2023). Of note was the greater hardseededness of cv. Tammin. The annual medics, yellow serradellas (apart from Yellotas) and Casbah biserrula were the most hardseeded annual legumes, which is in general agreement with the studies of Crawford et al. (1989), Loi et al. (2005), Revell et al. (1998, 1999) and Nichols et al. (2012).

Patterns of seed softening

The timing of seed softening over the summer–autumn period for seeds and pods placed on the soil surface was highly species-specific. The subterranean clovers softened rapidly, with the majority of seed softening occurring by late March, with little further softening beyond this date. This is in accordance with previous studies by Norman et al. (2006), Loi et al. (1999) and Newell et al. (2023). Chapman and Asseng (2001) showed that subterranean clover in the low- and medium-rainfall environments of WA was particularly vulnerable to germination from false breaks in March and April, because reliable winter rains do not generally commence until May. The similar patterns of Paradana and Frontier balansa clovers, Nitro Plus Persian clover, Cefalu arrowleaf clover and Electra purple clovers suggests that these cultivars are similarly vulnerable to false breaks. In contrast, several cultivars, including Scimitar and Santiago burr medics, SerraMax and Avila yellow serradellas, Casbah biserrula, Prima gland clover, Bartolo bladder clover and Sothis eastern star clover, demonstrated a more sigmoid pattern of seed softening, with a delay in the timing of major seed softening, making them less vulnerable to false breaks. This was particularly noticeable at Perth and accords with previous studies of Taylor (1996), Revell et al. (1998), Loi et al. (1999), Howieson et al. (2021) and Nutt et al. (2021a). The timing of the season break varies across southern Australia and is changing as a result of climate change (Flohr et al. 2021). Potential exists to match hard seed breakdown patterns with the season break at different locations.

The situation is different for pasture legumes used in permanent pastures in more reliable, higher-rainfall regions. Here, subterranean clover that germinates following favourable early seasonal rainfall can compete with other pasture components and weeds and deliver a long, productive growing season. However, subterranean clover seed softening patterns must still provide some protection against seedling loss after a false break and the ability to regenerate after a delayed break to the growing season. Although there is a diversity of seed softening patterns available, it is seemingly difficult to select pasture legumes that will have an optimal seed softening and germination strategy for the majority of seasons in these higher-rainfall environments. An alternative option may be to sow a mixture of legumes with differing seed softening attributes that are able to respond to different seasonal conditions.

Effects of shallow seed or pod burial

A reduction in subterranean clover seed softening from shallow burial was previously observed by Taylor and Ewing (1996) among early flowering ssp. subterraneum cultivars at a low-rainfall site in WA. This study showed a similar reduction in a wider range of cultivars at higher-rainfall sites that was consistent among the three subspecies. The contrasting increase in seed softening from pod burial of the yellow serradellas confirmed the results of Revell et al. (1998, 1999), Taylor and Revell (2002) and Howieson et al. (2021). This was particularly pronounced for SerraMax in Perth and Adelaide, which had rapid seed softening following pod burial, compared with little softening when pods remained on the surface. Seed burial of Prima gland clover also increased seed softening at Perth, being similar to the findings of Howieson et al. (2021). The annual medics were little affected by burr burial, conforming with the results of Taylor and Ewing (1996). Loi et al. (1999) found that shallow burial of Casbah biserrula seeds increased rates of seed softening. This effect was observed at Adelaide, but not at Perth or Canberra. Howieson et al. (2021) found that softening of Margurita French serradella increased with shallow burial, but this was not supported in our study. New findings in this study were reductions in seed softening at Perth following burial for Electra purple clover, Sothis eastern star clover, Cefalu arrowleaf clover and Nitro Plus Persian clover. The effects of burr burial among sites for Casbah biserrula, Bartolo bladder clover and the balansa clovers Paradana and Frontier were inconsistent and are discussed below.

Site effects on hard seed softening

This study confirmed the rate and extent of hard seed softening in annual pasture legumes is site dependent. Seed softening in general was greatest at Perth and least at Canberra, with softening being intermediate in Adelaide. Site differences have previously been shown for subterranean clover (Norman et al. 2006), balansa clover (Nair et al. 2004) and several other annual legumes (Howieson et al. 2021). Differences in the extent of seed softening among the three sites can largely be explained by differences in mean summer–autumn temperatures, which were greatest in Perth and lowest in Canberra, with Adelaide being intermediate. The warmer autumn temperatures in Perth also meant that seed softening continued later into autumn than at the other two sites. These results conform with the two-stage seed softening process described by Taylor (1981, 2005), in which the first stage relies on high temperatures to degrade the hilum and lens, pre-conditioning the seed for softening, and the second stage relies on fluctuating temperatures to render the seed permeable to water. The modifying effect of shallow burial on the extent of seed softening in some cultivars was also presumably because of changes in the temperature regimes compared with those on the soil surface. This was apparent in the soil temperatures at Perth, in which mean autumn temperatures 2 cm below the soil surface were 1.6°C lower than on the surface, with a 3.8°C reduction in temperature amplitude. It is likely that similar temperature modifications from burial also occurred at Adelaide and Canberra. Further studies are needed to determine whether differences in the mean or amplitude of temperatures had the greatest effect on seed softening.

Cultivar × site interactions

An important finding from this study was significant cultivar × site interactions for seed softening of some key cultivars. This included differences in relative rankings among sites for residual hardseededness of Bartolo bladder clover, Margurita French serradella, Yellotas yellow serradella and Leura subterranean clover, and for each of the February–June sampling times of some cultivars, including Margurita and Paradana balansa clover. Although the relative responses to shallow burial were similar across sites for subterranean clover, French serradella and annual medics, there were relative differences for some cultivars among sites. In particular, Casbah biserrula, Bartolo bladder clover, Paradana and Frontier balansa clovers, Scimitar burr medic and Herald strand medic demonstrated inconsistent burr burial effects among sites. Howieson et al. (2021) also noted site differences for seed softening of buried seeds of Casbah, with much greater softening at three sites in NSW than in WA, and Newell et al. (2023) found site × cultivar differences among serradella cultivars at two contrasting sites in NSW.

Implications for selection of annual pasture legumes for local farming systems

Pasture legume cultivars in Australia have typically been developed regionally, often with little evaluation across Australia. The differences among sites in this study highlighted the risks of taking a cultivar and agronomic package from one region, with the expectation of similar performance in another region. This is also of importance in other regions of the globe where annual legumes are used in sown pastures, including the Mediterranean basin, Chile, South Africa, the United States of America and New Zealand (Porqueddu et al. 2016). This issue is exacerbated with some of the more recently commercialised species, which in many cases have had little evaluation in regional farming systems beyond their initial location of development. Thus, there is a need for regional seed softening studies to go along with other agronomic characterisation for local cultivar recommendations.

This study has shown that the effects of seed burial on seed softening rates is not always applicable for cultivars across site and climatic differences. This may mean that revolutionary new practices, such as summer sowing in the low and medium cropping districts of WA and NSW (Howieson et al. 2021; Nutt et al. 2021a), may need to be tested in new environments to be sure the expected outcome eventuates. Indeed, Howieson et al. (2021) and Nutt et al. (2021a) suggested that Casbah biserrula is better suited to summer sowing in NSW than WA, because of a better matching of its seed softening pattern.

The fitting of suitable cultivars adapted to permanent or semi-permanent pastures in cooler, temperate regions, such as the Northern, Central and Southern Tableland and alpine Monaro regions of NSW, requires attention, particularly because many of the newly commercialised species were developed for lower-rainfall, pasture–crop rotation systems. Late-flowering subterranean clovers are well adapted to this zone but the search for more P-efficient, acid-tolerant species with greater tolerance of false breaks (Kidd et al. 2020; Hayes et al. 2023) has shown both yellow and French serradellas as promising alternatives. However, Newell et al. (2023) showed French, yellow and slender serradella cultivars have substantially different seed softening patterns from that of subterranean clover in southern NSW, raising concerns about their adaptation to these environments. The diversity observed among those species in this study suggests that there is an opportunity to select serradella cultivars for these environments with appropriate seed softening patterns and residual hard seed levels. The effect of burr burial to reduce hardseededness presents new opportunities for the use of yellow serradella in permanent pastures, where high hardseededness is less important, although its implementation is problematic in such environments.

This study showed sufficient diversity to breed and select other pasture legumes with improved seed softening dynamics for different environments and farming systems. Diversity for seed softening patterns also suggests the selection for greater resistance to germination from false breaks. Examples include the higher residual hardseededness of Tammin and Margurita than other subterranean clovers and French serradellas respectively, and the lower hardseededness of Yellotas than other yellow serradellas. Heritability for hardseededness has been shown to be high for subterranean clover (Nichols et al. 2013), French serradella (Nutt et al. 2021b) and balansa clover (Nair et al. 2004), indicating the potential to breed cultivars of these species with differing degrees of hardseededness. Although Peck and Howie (2012) described selection to reduce hardseededness in annual medics, all annual medics cultivars in this study were hardseeded, reflecting their development for ley pasture systems (Nichols et al. 2012). However, Taylor (1996) found considerable diversity for hardseededness among 34 burr medic accessions, with soft seed in June ranging from 6.8% to 69.6%, with a range in the timing of summer–autumn seed softening. This indicates the potential to breed annual medics with different levels of residual hardseededness and the timing of softening for different uses (e.g. ley or permanent pasture) and locations.

Study limitations and further research

This study produced empirical results on differences in hardseededness among genotypes and environments, rather than mechanistic ones. Soil temperature measurements on the surface and 2 cm below the surface at each site would have provided stronger distinctions among the sites and allowed temperature effects at the two depths to be examined in more detail. Seed moisture content has been shown to be important for development and softening of hard seeds (Taylor 2005; Harrison et al. 2021); however, it was not measured in this experiment. Measurement of seed moisture contents during seed softening should be examined in further studies to help explain genotype, site and soil depth effects. Furthermore, this experiment was conducted only over a single summer–autumn period after seed maturation and studies over subsequent summer–autumn periods are needed to measure longer-term seed softening.