Material was collected over two consecutive years from ten established populations of C. helmsii in northwestern Belgium (Fig. 1; Table 1). All sites were protected as nature reserves and/or have their habitats designated under the EU Habitats Directive (mainly oligotrophic to mesotrophic standing waters). For each population, the species was harvested by hand from a representative location where it formed dense mats and appeared to be flowering.

Plants were collected in October (November) 2024 and then stored at low temperature (6°C) for 66–86 days, after which flowers were isolated from the plants using tweezers.

As a substrate, we prepared a 1:2 mix of potting soil and sand, the latter of which was sterilized by heating it for 24 h at 120°C. The substrate was put in plastic trays (16 × 11 × 6 cm). Each tray had a wool felt pad at the bottom, was perforated, and was placed inside a larger tray to facilitate the uptake and drainage of excess water. Twenty flowers were sown per tray and buried very superficially. The trays were then subjected to a 14 h light / 10 h dark regime at 22°C (± 0.57, standard deviation) and 15°C (± 0.59), respectively. Temperature data are based on measurements from a logger. A two-hour transition between day and night has been omitted from the calculations. The lamps were LED tube lights (42 W, 6500 K). The trays were regularly watered, and their positions changed every four weeks.

The trays were checked for seedlings for 127 days. Seedlings were removed upon detection. When in doubt, a needle was placed next to a developing plant until confirmation. There were four trays (replicates) per population (Table 1). Eight trays with only substrate acted as controls; none of these showed germination except for a single graminoid specimen.

Additionally, 20 flowers from each population were dissected to determine the number of seeds per flower. We denote the mean number of seeds per flower for population n as µ_Sn, which was then used to estimate the actual number of seeds sown in the trays (i.e. 20 × µ_Sn).

Plants from each of the populations were sampled in October 2025 and then stored at low temperatures (fridge; 4°C) for 112–122 days, after which flowers were isolated from the plants.

Seeds were extracted from flowers under a stereoscope and transferred to 9 cm Petri dishes containing pure agar (VWR product code: MERC1.01614.1000). Given prior experience with C. helmsii stock indicating that microbial growth (fungi, bacteria) can be rapid and severe, and considering the relatively low number of seeds available, we opted to increase the number of replicates (dishes) and space the seeds adequately apart. Consequently, five seeds were sown per dish. While this number is lower than typical for germination tests (Poschlod et al. 2026), it was sufficiently robust for the intended (simple) analysis. The number of dishes per population are included in Table 1.

The dishes were subjected to a 14 h light / 10 h dark regime at 20°C (± 0.35) and 15°C (± 0.81), respectively (logger data, as above). The same lamps as in the previous experiment were used. The positions of dishes were changed at every scoring occasion.

Germination in the petri dishes was scored for a minimum of 65 days. A fraction of the seeds was removed at the earliest scoring times due to sanitary concerns arising from rapid microbial growth. Germinated seeds (seedlings) were removed upon discovery; a seed was considered germinated as soon as distal plant parts (root tip or cotyledons) emerged from the seed coat. Monitoring of the dishes was discontinued if no new germination had been observed in any dish from the same population for the past 14 days.

Each remaining seed at the end of the trial was evaluated under a stereoscope by longitudinal sectioning and assessment of the inner tissue contents. Indeed, non-germinated seeds vary fundamentally (Poschlod et al. 2026): they can either be viable (with an intact embryo, capable of germination but not yet germinated) or non-viable (incapable of germination, whether dead, empty, or containing an aborted or malformed embryo).

We analysed the proportion of seeds germinated (odds of germinated vs non-germinated seeds) using a generalized linear model (GLM) with a binomial distribution and logistic link function. Population was included as the sole explanatory variable (fixed effect), as to illustrate variation among the sampled locations. However, as our populations represent a random sample from a broader set of C. helmsii populations, and our primary objective is to gain a general understanding of germination rates (rather than, for example, differences among specific populations), we also fitted a generalized linear mixed model (GLMM) with population as a random effect. This allowed us to estimate the overall mean germination probability and the cross-population variation. Population BE07 was excluded from these analyses due to insufficient seed availability (Table 1).

For the GfF trial, germinated vs non-germinated seeds were equated to the number of observed seedlings vs the estimated number of non-germinated seeds in a tray. Since this latter number relies heavily on the estimated number of seeds per flower (µ_Sn), we explored the sensitivity of the estimate to this parameter by refitting the model using variations in µ_Sn (0.1–0.9 × µ_Sn). Indeed, since not all seemingly healthy seeds are necessarily viable, µ_Sn could theoretically be an overestimate (and the germination rate an underestimate).

For the GfS trial, germinated vs non-germinated seeds were equated to the germinated seeds (observed directly) vs the seeds sown. As a complement, we repeated the analysis by excluding the seeds that had either been removed early due to sanitary risks or were identified as non-viable during post-trial evaluation. This ensured this analysis was limited strictly to viable seeds.

Analyses were performed using R v.4.6.0 (R Core Team 2026) in RStudio v.2026.06.0 (Posit Team 2026), involving, among others, the R packages tidyverse v.2.0.0 (Wickham et al. 2019), stats v.4.6.0, lme4 v.2.0-1 (Bates et al. 2026), and emmeans v.2.0.3 (Lenth et al. 2026).

The number of seeds per flower varied strongly, with the majority (60%) having no developed seeds, and about one in four containing a single seed (1 seed: 24%; 2 seeds: 12%; 3 seeds: 4%).

Germination could only be detected at a relatively late stage, when at least the cotyledons were fully developed (Figs 2, 3). For the ten populations combined, a total of 67 seedlings were recorded across the 40 trays (800 flowers, equalling an estimated 500 seeds). Ninety percent of the seedlings were recorded within 75 days (Fig. 4).

The fixed-effect model estimated germination rates ranging from 6.8% (± 2.7% s.e., standard error) to 43.8% (± 12% s.e.; Fig. 5). The random-effect model yielded an overall mean germination probability of 13.8%. Accounting for variation among populations, this model estimates that 95% of C. helmsii populations have germination probabilities ranging from 4.8% to 34.1%.

The sensitivity analysis is shown in Fig. 6. The estimate of the overall germination rate proved fairly robust to variations in µ_Sn, in that conceivable misestimates of the number of viable seeds per flower (~0.9–0.6 × µ_Sn) still yielded low germination rates.

Germination was clearly observable in the agar plates, following an expected pattern (Figs 2, 3). During germination, seeds swelled slightly, the radicle emerged from the seed coat, fine root hairs developed, the hypocotyl elongated, the cotyledons unfolded, and the green intensity of the upper plant half increased. In a few cases, the cotyledons were the first organ to rupture the seed coat. The developing seedling either completely detached from the seed coat or the seed coat remained attached to the tip of one of the cotyledons. Notably, the radicle apices consistently exhibited a reddish coloration.

Germination was observed more rapidly than in the previous experiment (90% within 40 days; Fig. 4). Although filamentous fungi rapidly developed, they eventually receded, and germination did not appear to be hindered. Later observations also detected microbial growth with a more mucilaginous habitus, sometimes associated with seeds. However, it remains unclear whether this had a neutral or potentially negative effect on germination.

From the 333 sown seeds for the ten populations combined, 227 seedlings were counted. The fixed-effect model estimated germination rates ranging from 60.0% (± 7.7% s.e.) to 86.7% (± 8.8% s.e.; Fig. 5). The random-effect model yielded an overall mean germination probability of 68.5%. Accounting for variation among populations, this model estimates that 95% of C. helmsii populations have germination probabilities ranging from 57.1% to 78.0%.

Due to microbial growth (sanitary concerns), 46 seeds were promptly removed and consequently considered non-viable. Thus, a total of 60 non-germinated seeds remained at the end of the trial for detailed evaluation. The majority of these (50 seeds, 83%) were found to be non-viable (empty, or with incomplete or amorphous tissues). Only 10 seeds (17%) appeared fully intact.

Based on these more conservative data, the fixed-effect model estimated germination rates across populations to range from 88.9% (± 6.0% s.e.) to 100% (± 0%). However, due to the limited number of seeds, high germination rates, and low variation, the model now fitted poorly (residual deviance << degrees of freedom). Similarly, attempts to include population as a random effect failed due to singular fits. This only leaves the intercept-only (null) model, which estimated an overall mean germination probability of 95.7%.

Our results confirm that a significant proportion of C. helmsii flowers in the studied region contain seemingly healthy seeds (estimated at 40%). After simple cold storage, these seeds exhibited high germination rates (estimated at 68.5% for a given population), with subsequent analysis revealing that non-germinating seeds were essentially non-viable. When these were excluded, nearly all seeds successfully germinated (estimated at 95.7%). The small fraction (~5%) that failed to germinate may represent very late germinators, not captured by the experiment (but see below). There is variation among populations, though rates are of the same order of magnitude (and never zero).

When seeds were subjected to more challenging (realistic) conditions, i.e. germination from floral tissues in an organic/mineral bed, the germination rate dropped to 13.8%. Notably, applying the same analysis post hoc to the data from the analogous trial in D’hondt et al. (2016) yields an estimated rate of 11.3%. The GfS trial clearly demonstrated that a proportion of the seeds we initially deemed healthy were, in fact, non-viable (roughly one in three). This suggests that the germination rate from the GfF trial is an underestimate. However, the sensitivity analysis revealed that, for a similarly high germination rate as observed in the GfS trial, the true number of viable seeds per flower would need to be as low as one-tenth of our observed count. This is implausible, leading us to conclude that the lower germination rate is real and not an artifact. We can thus speculate about processes—pre- or post-germination, lethal or non-lethal—that may cause the lower emergence: e.g. a lack of light for germination (Milberg et al. 2000), or fatal germination at depth (Fenner and Thompson 2005). At the same time, although the GfF trial more closely approximated in situ conditions, some complexity remained neglected. For example, it is plausible that seed overwintering or spring germination under potentially submerged conditions could affect recruitment (positively or negatively). We suspect that the slower pace observed in Fig. 4 for the GfF trial reflects a time lag in the detection of germination, rather than germination itself.

Crassula helmsii germinated rapidly under conventional in vitro conditions. The observations suggest it to be a positively photoblastic species (i.e. requiring light to germinate). The red radicle tips are likely caused by anthocyanins, which are known in Crassulaceae (Molisch in Thakur and Nozzolillo 1978; Góraj-Koniarska et al. 2015), although it remains uncertain whether this is a regular or stress-induced phenomenon (e.g. due to intense light). From a methodological perspective, the use of an agar substrate proved effective. The seeds are most easily handled using moistened needle tips. Microbial growth ultimately proved less problematic than initially anticipated. In fact, in the context of our trials, we conducted a small sterilization test using excess seeds from three populations, where seeds were placed on filter paper with diluted bleach and polysorbate for 15 minutes (Niko Robberechts unpubl. data). This treatment had no significant effect on germination success, though visually, the microbial load on the plates appeared reduced. Complete sterilization with bleach (e.g. by submersion) remains practically challenging for these small seeds. Therefore, if sterilization is desired, it is advisable to consider alternative methods, such as heat (Godefroid et al. 2017) or UV treatments (Kamel et al. 2022).

All plant material underwent a period of cold storage representing simplified winter conditions. The observation of near-complete germination thus suggests that C. helmsii seeds are either non-dormant or, more likely, physiologically dormant with cold conditions (“chilling”) releasing dormancy (Poschlod et al. 2026). Indeed, physiological dormancy is considered the standard type within the Crassulaceae family (Baskin and Baskin 1972, 2014; Martínez-Villegas et al. 2012). It is important to recall that seeds with physiological dormancy—unlike all other dormancy types—can switch back and forth between dormant and non-dormant states in response to environmental conditions (Finch-Savage and Footitt 2017). For example, the acquisition of secondary dormancy has been demonstrated in Sedum pulchellum Michx. (Crassulaceae), which moreover combines dormant and non-dormant seeds within the same plant (Baskin and Baskin 1977). In fact, acquired secondary dormancy might also have contributed to the lower emergence in the GfF trial, and the viable but ungerminated fraction in the GfS trial might actually have consisted of dormant seeds (see above).

This transition from dormant to non-dormant also means that seeds with physiological dormancy are capable of forming seed banks. Indeed, among crassulacean species present in Europe, seed bank types are classified as transient in some genera (e.g. Hylotelephium H.Ohba, Phedimus Raf.), but as short- to long-term persistent in others, such as Sedum L. and Petrosedum Grulich (TRY database; Fitter and Peat 1994; Kleyer et al. 2008; Tavşanoğlu and Pausas 2018; Kattge et al. 2020). For Crassula L., these sources interpret the results on C. tillaea Lest.-Garl. from Levassor et al. (1990) as indicating short-term persistence, whereas its apparent presence in samples from unploughed soil over 40 years might also hint at long-term persistence (i.e. remaining viable for over five years; also see Peco et al. 2003).

Seed and dormancy traits are strongly conserved within families and genera, so the above likely also applies to C. helmsii (Willis et al. 2014; Seglias et al. 2018). The emergence of C. helmsii from sediment samples in Australian studies has already been interpreted as germination from the seed bank (Nicol et al. 2003; Nicol and Ward 2010). Spontaneous germination in natural populations of C. helmsii has not yet been confirmed, and it remains true that distinguishing seedling emergence from emergence via vegetative fragments is extremely difficult. The small size of the seeds and seedlings, combined with the lack of distinguishing features (e.g. cotyledon shape), makes visual confirmation of germination in the field virtually impossible. Nevertheless, there do not appear to be any fundamental barriers to in situ seed formation, dispersal, or seed bank formation, and so we caution that the absence of evidence for germination should not be mistaken for evidence of absence.

Seed formation also raises questions about the pollination ecology of C. helmsii. It is unlikely that seeds are formed apomictically (without pollination), but self-pollination (autogamy) is known within the genus (Thiede and Eggli 2007). The flowers appear to be entomophilous (Thiede and Eggli 2007), and visits by Syrphidae (hoverflies) have been documented within the species’ European range (Diaz 2012). Given the high density of flowering stems (D’hondt et al. 2016), geitonogamy is most likely, and our results thus suggest that C. helmsii is a self-compatible species. The population-level implications, e.g. with respect to genetic diversity and adaptive potential, remain to be determined. As Green and Noakes (1995) demonstrated in animals, even a small component of sexual reproduction can be highly advantageous in an otherwise clonal life cycle.

Recruitment via seeds has varied implications for management, which is significantly complicated. Prevention must account for the potential presence of seeds, whether targeting pathways (e.g. as stowaways in the water plant trade; Van den Neucker and Scheers 2022) or in the field (e.g. through the use of filters; Bok et al. 2025). Field interventions aimed at plant removal must be executed with high precision to eliminate any potential seed bank, while also ensuring biosecurity (van der Loop et al. 2022). Approaches that accept the presence of C. helmsii but seek to suppress its dominance through biotic interactions appear more cost-effective in this regard, e.g. by introducing plant competitors (van der Loop et al. 2023; Gaudichet et al. 2026), or specialized herbivores (Varia et al. 2022). However, even here, it must be considered that C. helmsii possesses a life stage that can at least temporarily evade stressors.

Plant species from aquatic environments often exhibit a narrow syndrome that combines asexual (clonal) and sexual traits, which partly explains their frequently large ranges (Santamaría 2002). The findings of this study suggest that established C. helmsii populations in Western Europe also display this combination, as viable, easily germinable, and likely physiologically dormant seeds are readily found. Further research is needed to unravel the exact role of generative versus vegetative reproduction in (meta)populations of this highly invasive species. Potential avenues could include further seed bank studies using soil samples, or small-scale population genetic studies to assess diversity patterns and distinguish ramets from genets (Kaljund et al. 2013; Holt et al. 2020; Loh et al. 2020).