Seed-based ecological restoration is becoming an increasingly popular and cost-effective approach to re-establish native vegetation in a degraded site (Larson A. J. S. et al., 2023). It relies on using species-rich seed mixes that represent the vegetation community of a reference native ecosystem (Gann et al., 2019). For seed-based restoration to be performed at scale, the local native seed supply chain should be able to provide seed in the quantity, quality and species diversity required (Pedrini and Dixon, 2020). Unfortunately, in most scenarios, seed availability is not enough to meet demand, the quality is poor or unknown (Gibson-Roy et al., 2021), and in areas where the supply is overreliant on direct seed collection from natural populations, the cost can be prohibitive (Brancalion et al., 2019; Pedrini et al., 2022). Even when seeds are available, seed delivery to the site needs careful planning, site preparation, and impeccable timing to maximise the chance of successful seedling emergence and plant establishment (Shaw et al., 2020).

Various seeding approaches have been tested for delivering seeds to site (Masarei et al., 2019). Precision seed delivery (seed drill), whereby seeds are incorporated into the soil, has usually proved more effective than broadcasting seeds on the soil surface (Svejcar et al., 2022). Precision seeding equipment and methods are often derived from agricultural equipment developed to handle a specific type of seeds (e.g., wheat, canola) and, therefore, struggle to handle the high variability in size and morphology of native seed mixes. A high degree of customisation of seeding machines can account for such complexities, but that often comes at the expense of seeding speed and scale (Shaw et al., 2020). Moreover, accommodating the variability of a native seed mix in a single seeding device necessitates compromises that might favour seeds with certain traits over others.

For example, during transport and seeding, vibrations can cause seeds to be separated (segregation) based on size and density, with smaller and denser seeds ending in the lower part of the hopper (Liu et al., 2013; Rosato et al., 2002). This would then result in an uneven distribution of species across the site. Using bulking agents, such as vermiculite or sand, could overcome this issue but further increase logistical complexities, including sourcing the material, delivering it to the site and ensuring constant mixing with the seeds, which would slow down seeding operations.

Another approach to improving seeding efficiency is directly modifying seed shape and size to reach a more homogenous mix. This is often achieved via seed processing by removing (Barberis et al., 2023) or reducing (Berto et al., 2020; Guzzomi et al., 2016) appendages (e.g., awns and hairs in grass florets) that can impact seed flow (e.g., bridging). A further step involves applying fine powders (fillers) and binders to the seed to modify its shape, size and density via seed coating (Brown et al., 2021; Halmer, 2008; Pedrini et al., 2017; Taylor et al., 1998). There are different kinds of seed coatings, and the one that modifies the shape and size of the seed unit the most, seed pelleting, can be effective in improving seed delivery efficiency, especially in small-seeded species. (Gornish et al., 2019; Hoose et al., 2019).

Like seeding equipment, such technology originated in the crop and horticulture seed industry and has been developed to pellet one species at a time (Afzal et al., 2020). When applied to native species for ecological restoration, most studies have performed seed pelleting on individual species (Beveridge et al., 2020; Hoose et al., 2019; Madsen et al., 2014; Pearson et al., 2019; Pedrini et al., 2023; Turner et al., 2006; Westbrook et al., 2023). This allows for species-specific customisation of pelleting recipes, accounting for factors such as germination requirements, which may improve establishment outcomes. However, this approach presents some practical drawbacks when applied to diverse seed mixes. A recent study by Pedrini et al. (2023) highlights that the time required to pellet seeds of 15 small-seeded Myrtaceae species ranges between 37 and 188 min per species, with an average of 98 min. This means that the more diverse a mix is, the longer and more expensive the pelleting will be, making the process economically non-viable, thus nullifying any potential increased efficiency of seeding with pelleted seeds.

This paper describes a novel approach that overcomes such logistical issues by pelleting seeds of multiple small-seeded species in a single pass while maintaining singulation (one seed per pellet), thus potentially reducing the time and cost associated with seed pelleting.

The first part of the methods section (4. 1 Multispecies pelleting method) provides a detailed explanation of the process.

In the second part of the methods section (4.2 Method testing and validation) and the results section, two case studies of the multispecies pelleting approach (MSP), of 13 and 19 species, are presented and compared with a model that estimates the time and cost that would have been required if those species were pelleted using the traditional single-species method (SSP). Such a model was built using data from 32 pelleted seed batches belonging to 28 species of comparable size and traits to the ones tested with the MSP.

This preliminary analysis comparing the time and cost associated with both methods provides a general indication of the potential efficacy and impact of the MSP approach.

Seed pelleting, defined as the process of adding material to a seed until its original shape is no longer evident (Pedrini et al., 2017), is generally performed by alternatively applying a liquid binder and a fine powder (filler) until the seed is covered, the predetermined amount of pelleting material is fully used, or the desired pellet size is reached.

The most commonly used equipment for pellet seeds are rotating pans (Scott et al., 1997) and rotary coaters (Halmer, 2008).

As the name suggests, a rotating pan consists of a circular pan with raised sides to retain seeds and material inside the pan, similar to a cement mixer. The rotating motion allows the seed mass to be constantly mixed while sitting on the lower part of the pan, ensuring even distribution of pelleting material among all the seeds. The friction exercised by the pellets tumbling on each other results in the compaction and shaping of pellets into approximately spherical shapes (Bennett and Lloyd, 2015; Gregg and Billups, 2010).

A rotary coater (or rotor-stator) consists of a vertical metal drum (stator) with a rotating concave disk (rotor) at its bottom. Seeds are placed in the drum, and the bottom disk’s rotation causes the seeds’ mass to rapidly move along the lower side of the wall. Material is then added to the mass of moving seeds. Deflectors mounted on the inward-looking side of the drum ensure proper mixing of seed and material. A constant flow of air is usually forced in the gap between the edge of the disk and the wall of the drum to avoid material and seeds falling through the gap.

The rotary coater is often preferred for native seeds (Pedrini et al., 2020) as it allows for faster operations and is generally easier to operate. However, it can present problems with very small-seeded species, as they can either get stuck in the small gap between the drum or be blown out of the pan by the airflow. A rotating pan was preferred in this study because we focused on pelleting small-seeded species.

The machine used was a PC-S rotating 340 mm segmented pan (Hoopman, Aalten, NL), connected to a peristaltic pump and compressed air line for precision delivery of liquid binder via spray gun mounted on an adjustable arm (Figure 1). Pan rotation speed and liquid feed rate from the peristaltic pump can be adjusted from the control panel. Spray intensity and liquid delivery rate can be fine-tuned directly on the spray gun by rotating the front nozzle and nob on the back of the gun, respectively.

The liquid binder used across all batches was an 8% solution of Polyvinyl alcohol (Glow Paint Industries, Urangan, QLD, Australia) without colour for the internal layer and with a dye (SATEC, Elmshorn, Germany) for the external layer. The powder used for the internal layer was micronised Azomite^® (Dr Greenthumbs, Bellambi, NSW, Australia) and talcum powder was used for the external finishing layer. Azomite was chosen for ease of pelleting (regular pellet shape), relatively high density, and limited impact on germination (Pedrini et al., 2023), while finishing the talc layer was selected to improve pellets’ mechanical resistance, reduce azomite dust-off and allow even and easier colouring of the final layer. The powder was manually dusted on the seed with a paintbrush or strainers. A set of sieves of mesh size ranging from 0.25 mm to 2.36 mm was kept next to the pelleting station for sorting pellets during and after pelleting operations. A dehydrator (Sunbeam, Boca Raton, FL, United States) is also needed to dry the seed after the pelleting process.

Such materials were selected based on previous experiences (Pedrini, 2019) and published protocols (Pedrini et al., 2023) but are just one of many potential material combinations. Material selection should be informed by availability, cost, the desired pellet properties (e.g., density, mechanical resistance, water permeability), and the potential effect of the material on seed germination and seedling establishment.

This section provides a detailed explanation of the preparatory, pelleting, and post-pelleting steps. Ideally, such instruction would guide researchers and practitioners in testing the MSP method, acknowledging that customisation and improvement might be needed to address specific needs and equipment availability.

The average seed weight of a batch before pelleting was 3.82 ± 0.28 g and was increased to 77.45 ± 18.09 g after pelleting for an average increase of 18.58 ± 3.49 fold (xWeight). The average volume per batch was 6.31 ± 0.49 mL before pelleting and was increased by 16.38 ± 3.09 fold (xVolume) to 94.51 ± 19.89 mL. The average weight of a thousand pure seed units was increased by 18.16 ± 3.18 times (xTPSU) on average, from 0.76 ± 0.14 g to 6.27 ± 0.43 g. The average duration of the pelleting process, without accounting for 30 min of pre and post-pelleting operations, was 87.66 ± 9.06 min, ranging from a minimum of 23 min for Gastrolobium capitatum, to a maximum of 210 min for Calothamnus quadrifidus (see Supplementary Table S2).

When pelleting parameters were plotted on a linear model against the pelleting duration (min), the best fit was identified for Thousand Pure Seed Unit weight increase (xTPSU, R ²: 0.702), followed by weight increase (xWeight, R ²: 0.641) and volume increase (xVolume, R ²: 0.583) (Figure 4).

The MSP from the Avon Wheatbelt bioregion (AVW) batch consisted of 13 species for a total seed weight of 57.08 g at an average TPSU of 1.186 ± 0.46 g and an estimated 107,750 Pure Live Seeds. Seeds were pelleted for 362 min, using 790 g of liquid binder and 1,810 g of powder until most pellets reached the targe size range of 2 mm–2.36 mm, for a total weight of 1,118 g and a TPSU of 9.376 g. When quality tested, 4% of the pellets were empty (dead balls), 76% contained one seed, and 20% contained two or more seeds (Figure 5). An estimated total of 115,616 pellets were produced, 110,994 of which were filled.

The MSP batch from the Esperance Plains bioregion (ESP) of 68.67 g comprised 19 Myrtaceae species at an average TPSU of 0.267 ± 0.05 g and an estimated total of 200,000 PLS. The pellet size range was set between 1.6 mm and 2 mm and was reached in 318 min using 842 g of liquid binder and 1,868 g of powder. The final weight of the pelleted batch was 1,117 g at a TPSU of 5.302 g. Tests on the pellets revealed 8% of empty pellets, 78% of single-seed pellets, and 14% of agglomerates. The total number of pellets produced was 210,659, of which 193,807 were filled.

Assuming that the material and seeds used would have been comparable in both SSP and MSP, the difference in cost between the two approaches is due to the differences in total time (preparatory, pelleting and post-pelleting operations) and associated labour costs. For the 13 species of the AVW mix, MSP lasted 403 min, and for the 19 species ESP mix, it was 363 min.

To estimate the time that would have taken to pellet each seed lot individually with the SSP method, the equation for the linear regression of xTPSU against pelleting duration (Time) was used. T i m e   ⁡ min ⁡ = 2.39   ·   x T P S U + 44.3

The AVW mix time was estimated at an average of 177 ± 30 min per species for a combined total of 2,298 min. The average time for the ESP mix was 195 ± 21, and the sum was 3,700 min (Table 1).

In the AVW mix, the MSP approach allowed for a 5.6-time reduction in time compared to SSP, and resulted in a 3.1-fold reduction in the total cost of the pelleted seed batch. This difference was more accentuated in the ESP mix with a 10.2-fold reduction in pelleting time and consequent 3.9-fold reduction in total cost.

When considering the cost on a seed number basis, using the cost of a thousand pure live seeds (TPLS $) in the AVW mix, the cost for pelleted seeds compared to untreated seeds was doubled for MSP but 6.2 times higher for SSP. For the ESP mix, MPS had a 1.5 increase in TPLS $ compared to a 5.9 increase in cost for SSP (Table 1).

While experimental seed enhancement technology might be promising, it needs to be scalable and cost-effective for its adoption by seed suppliers and restoration practitioners (Brown et al., 2021).

The Multi-Species Pelleting method (MSP) presented in this study demonstrates potential time and cost savings compared to the traditional Single-Species Pelleting method (SSP).

Such savings become more accentuated as the number of species to be pelleted increases, as seen in the AVW mix of 13 species and the ESP mix of 19 species.

However, MSP is still responsible for a considerable cost increase compared to untreated seeds.

To address this issue, two approaches for further development and experimentation are suggested: 1) Find approaches to improve the MSP method efficiency and scalability, and 2) evaluate the benefit of pelleted seed-based restoration compared to traditional methods to ensure that the increase in seed cost is offset by better seeding efficiency and plant establishment.

The relatively long time (between 5 and 6 h per mix) required to pellet the two MSP batches was largely due to the fact that these were the first-ever mixes tested with this approach, and extra care was taken to avoid issues during the process. However, as the seed pelleting operator gains experience with this approach, the duration could be reduced to times similar to those of SSP, further enhancing the cost efficiency of this approach. This trend could already be observed, with the second attempt on the ESP mix in 2023 being 44 min faster than the first batch in 2022. Such time saving was achieved even though the number of species and quantity of seeds were higher on the second attempt. This element is particularly relevant for further scalability of the process, as the time required for pelleting is not appreciably affected by the number of species in the mix or batch size.

A method to improve speed and capacity is to couple the pan coater with a rotary coater and run a pelleting session across two machines. The rotating pan is required to pellet seeds in the first phases (<1 mm), as it allows more control of the variables and reduces the risk of accidentally losing small seeds; however, when pelleted seeds reach a relatively large size, (>1 mm), further size increase to reach the final dimension can be slow. Moving the partially pelleted seed mix to a rotary coater would allow for a much faster second accretion phase, reducing the time required to complete an MSP batch. Moreover, if the rotary coater is equipped with an automatic and programmable powder and liquid delivery system, this could be set up to autonomously complete the process, allowing the operator to commence a new batch on the rotating pan.

A limiting factor is, however, the total size of the seed mix that can be processed in one session. The pelleting pan used in this study is equipped with a 350 mm pan that allowed for the pelleting of a total estimated 200,000 PLS for the ESP mix, which was close to capacity towards the end of the process. This limit can be raised by installing a larger pan, such as a 600 mm one (Figure 1), that could increase the number of seeds that can be pelleted in one session to more than a million.

Assuming that with the improved approach, we can obtain a 1 million pelleted PLS mix in one session, that the average seed-to-established plant conversion is 10% (James et al., 2011), and that the target plant density is 10,000 per Ha, a single batch would be sufficient to cover a 100-ha restoration site. By employing large-scale agricultural seeding equipment (e.g., canola seeder), such an area over a flat terrain could be seeded in a matter of days, rather than the weeks required with current restoration seeding methods (Pedrini et al., 2023). This would further reduce seed-based restoration costs and allow for seeding operations to occur when soil and environmental conditions are optimal for seedling emergence and establishment, maximising the chances of seed-based restoration success, especially in arid and semi-arid environments (Shackelford et al., 2021).

Clearly, such assumptions need to be field-tested, and a more comprehensive cost-efficiency analysis based on the cost per successfully established plant, other than just the cost of seeds, needs to be performed. Such studies, in different restoration scenarios and species mixes, would then allow for better-informed decisions on whether MSP is a viable option and to what extent it can improve seed-based restoration efficacy.

This study has focused exclusively on the logistical improvement afforded by seed pelleting. However, seed germination and successful seedling establishment are ultimately the key outcomes of seed technology. In this study, I did not evaluate seed performance as it was previously assessed for the SSP species and method by Pedrini et al. (2023), and, on average, pelleting showed no detrimental effect on seedling emergence. To further increase the chances of germination and successful plant establishment, active compounds, such as predator deterrents (Pearson et al., 2019; Taylor et al., 2020), protectants from diseases and herbicides (Brown et al., 2021; Hoose et al., 2022), and survival, germination, and emergence promoters (Erickson et al., 2017; Larson J. E. et al., 2023; Pedrini et al., 2021) can be incorporated with the pelleting material (Madsen et al., 2016).

New creative approaches have recently been developed to introduce seeds of native species within agroecosystems using existing agricultural seeding equipment. For example, Westbrook et al. (2023) developed an innovative multi-seed moulding method to agglomerate seeds of Common milkweed (Asclepias syriaca L.) in the shape of corn seeds (Zea mais), that can be seeded by farmers alongside the crop and support the conservation of monarch butterfly, by introducing in farmland its main food source. A similar approach was developed by Corli et al. (2024) by encrusting rice seeds with the micro-seeds of endangered Linderia procumbens, allowing farmers to reintroduce a conservation priority annual species within their crop, with no impact on crop yield.

Even though this analysis focused predominantly on cost reduction, there is a key element that is difficult to quantify but must be considered. Native seeds, when collected from a natural population, can impact the population’s reproductive capability, especially for annual and short-lived perennial species (Bucharova et al., 2023), or damage the surrounding vegetation by trampling or accidental introduction of diseases. Until native seed production in farm settings becomes a viable alternative to wild seed collection, the potential collateral ecological damages of collection should be acknowledged. Therefore, limiting seed wastage by maximising the use of such valuable resources becomes a moral imperative, even when the monetary cost for doing so might be higher.