Insights from a century of data reveal global trends in ex situ living plant collections

The meta-collection has reached peak capacity

To characterize the growth dynamics of living collections over time, we analysed all collections individually (Supplementary Fig. 1) but also combined all 50 living collections and reconstructed the contents of this meta-collection over a 100 year period (1921–2021). We found that growth in number of accessions adheres to a sigmoidal growth curve with distinct lag, accelerated, transitional and stationary phases (Fig. 1a). The meta-collection experienced its most rapid expansion between 1975 and 1992. Following this period, the pace moderated, peaking in 2008, then levelling off and entering a phase of gradual decline from 2015 through 2021. We explored the dynamics underlying stationarity by examining turnover. Here, we observed a lag between the increasing rates of accession gain and the increasing rates of accession loss (Fig. 1b). However, as the rate of incoming accessions drops off, the rate of accession loss continues, leading to a net loss of accessions for the first time in 2009 (Fig. 1b), defining the peak size of the meta-collection. In effect, this sigmoidal pattern signifies density-dependent growth of collections, where the rate of increase is influenced by the existing number of accessions, eventually reaching a saturation point where no further growth can occur due to limited resources. In other words, there is now zero to negative growth, as the meta-collection has reached its carrying capacity, an ecological concept that is commonly applied in the context of sigmoidal density-dependent growth curves¹³.

Announcement

The meta-collection has reached peak diversity

To understand how the diversity of the meta-collection has changed over time we analysed the accumulation rates at different taxonomic hierarchies. Within the meta-collection, all levels exhibited similar sigmoidal curves, with species, genera and family diversity all reaching stationarity (Extended Data Fig. 1a). We further quantified diversity by calculating changes in phylogenetic diversity (PD) and show that PD had similarly plateaued within the meta-collection (Fig. 1c). Notably, diversity measures reach stationarity earlier than total accessions. For instance, while the overall accession-level capacity of the meta-collection did not peak until 2007, PD had already levelled off by 1990, at a point when the meta-collection was only at 75% of its maximum size. Consequently, expanding the meta-collection from 75% to its maximum size, had a negligible impact on the overall capture of PD (Fig. 1c), with implications for efficiencies in the storage of diversity. Notably the mean number of individuals per species, a common proxy for intraspecific diversity, also reaches stationarity, consistent with the peak of carrying capacity (Extended Data Fig. 1a). That diversity metrics peak ahead of total accessions is in part due to nested taxonomic hierarchies, but other limiting factors may include taxonomic duplication within and among collections, constraints of horticultural expertise limiting the diversity that can be grown, and the temperate ecological niche biases operating across a predominantly global north network of botanic gardens¹.

The meta-collection has reached peak wild

We examined provenance, as a critical factor influencing the utility of living collections. We grouped provenance into four categories: wild-origin (sourced directly from nature), wild-derived (documented asexual or sexual propagation from a wild-origin accession), garden-origin (from cultivated collections without a traceable wild link) and unknown-origin. We then analysed how the proportions of these provenance categories changed over time (Fig. 1d). Results show a notable improvement in provenance identification, with unknown origins dropping from a rolling average of 60% to a still high level of 29% in 2021. Wild-origin accessions meanwhile increased from a rolling average of 6% to a current value of 29% in 2021 but peaked in 1995 at 34.1% (termed peak wild) (Fig. 1d). We explored the causes of peak wild, and by examining the provenance of annual introductions to the meta-collection, we observed that wild collections (but not other provenances) showed a remarkable immediate and pronounced downward trend post-1993 (Fig. 1e) ultimately resulting in a 44% reduction in the acquisition of wild-origin material. This decline is most easily attributable to the Convention on Biological Diversity (CBD), which came into force on 29 December 1993, and assigns sovereignty over biodiversity to national governments, creating potential and realized obstacles to the sharing of genetic materials. Given the inferred influence of CBD on wild collections, we then examined whether living collections were also accessioning less internationally sourced material and indeed found a congruent 38% decline in the accessioning of non-native plants post-1993, with a notable mirror increase in accessions of native plants (Fig. 1f). In effect, we find that the proportional and absolute decline in wild and international acquisition are off-set by recycling garden-origin and wild-derived accessions, and by greater accessioning of native species.

Numerous constraints limit ex situ conservation

We were curious whether the patterns observed across the meta-collection also extend to the International Union for Conservation of Nature (IUCN) Red-listed plant species threatened with extinction. To assess this, we retrospectively applied current IUCN (December 2023) designations to historical data and examined the growth curves. We found that threatened collections display the same sigmoidal curve as the meta-collection as a whole, indicating that threatened species accumulation is similarly constrained (Fig. 2a). However, threatened collections have not yet fully plateaued and lag behind the trends of the meta-collection as a whole (Fig. 2b and Extended Data Fig. 1b). Interestingly, we saw no visible increase in the rate of accumulation of threatened species with the onset of IUCN designations post-1978 (Fig. 2a). So, we then asked what factors might be constraining the acquisition of threatened species. One obvious constraint is that threatened plants are by definition rare in the wild and, all things being equal, are less likely to be accessioned into living collections. To explore this, we examined the pattern of endemics versus widespread species and found a similar lag pattern for endemics, suggesting that rarity in the wild is a limiting factor for inclusion (Fig. 2b). Importantly, the distinct patterns for threatened species and endemics are not the result of poorer survival rates within collections, as addressed in the subsequent section exploring median survival rates. Using the same dataset with retrospectively applied current IUCN designations we then asked whether the provenance trends observed for the meta-collections as a whole, also applied to the threatened species pool. Here, we found similar trends to that seen for the meta-collections as whole; that is, a clear decline in wild accessions post-1993, and an increasing proportion of threatened plant accessions being recycled as garden-origin or wild-derived post-1993, potentially limiting the accrual rate of new threatened species (Fig. 2c). Given little to no evidence that extra effort is being taken to accrue threatened species at the level of the whole meta-collection, we then sought to understand responsiveness of the meta-collection to threat designation. Here, it is necessary to distinguish between the intentional acquisition of threatened species versus the species that are already in the collections and simply acquiring newly designated threat status. We first compared the rate of threat designation against the rate of addition of threated plants to the meta-collection, noting that extinction risk is accruing at a much faster rate than inclusion in the meta-collection (Fig. 2d). Limited acquisition of threatened species is therefore not due to a limiting pool of threatened species. To exclude the effect of already accessioned species accruing new threat designations, we first examined the relative proportion of threatened to non-threatened species that were newly accessioned each year (1978–2021) and found a very limited increase in the proportion of threatened accessions being newly incorporated (effectively a 1% increase over 40 years) (Fig. 2e). We then investigated whether the number of species added to the meta-collection significantly increased in the 10 years after threat designation (mean = 232.1, s.d. = 42.3) versus 10 years before threat designation (mean = 216.4, s.d. = 48.8), but found no significant effect (P = 0.45). Together these two analyses confirm that a lack of response to threat designation contributes to slow accrual rates (Fig. 2f). Finally, we examined the relationship between numbers of species and numbers of individuals per species as this trade-off may constrain the accrual of threatened species. We compared the number of individual plants accessioned for threatened versus non-threatened species, as a proxy for genetic diversity, and found that threatened species tend to have more individuals than non-threatened species (Fig. 2g), consistent with conservation value of intraspecific genetic diversity. However, it is evident that the capacity constraints on the meta-collection as a whole are also impacting on the conservation of intraspecific diversity as the mean number of individuals per threatened species has also reached stationarity (Fig. 2h).

Announcement

Announcement

a, Proportion of threatened versus non-threatened species within the meta-collection over time. b, Same as a with endemic versus widespread species overlayed for comparison. c, Number of new accessions of threatened species by provenance with lines equating to a 5 year rolling mean. d, Accumulation of species designated as threatened by the IUCN versus the accumulation of designated threatened species in the meta-collection (starting from the first appearance of digitally available IUCN Red List in 1978). e, Proportion of threatened plants accessioned relative to all new accessions in a given year. f, Rates of accessioning of threatened species in 10 years before designation versus 10 years after designation (year 0 is when a species was first designated as threatened). g, Number of individual plants per species for threatened versus non-threatened plants. h, Mean number of individual plants per threatened species over time.

Effective ex situ conservation leaves measurable signature

At the level of the entire meta-collection, accessioning of threatened species constitutes a small portion of total acquisition effort and this has not changed very much (Fig. 3a). However, it is vital to recognize that certain ex situ conservation programmes can exhibit very different behaviours, such as the International Conifer Conservation Programme (ICCP) which is run out of the Royal Botanic Garden Edinburgh. Here, we compared the ICCP against the meta-collection, across a range of parameters. First, we examined the proportion of threatened species within the ICCP versus the meta-collection as a whole and found a fivefold increase in the proportion of threatened species within the ICCP (Fig. 3b). Then we asked whether the ICCP exhibited an unusual signature with respect to provenance and found a 3.5-fold increase in wild collected species within the ICCP versus the meta-collection (Fig. 3c). We explored the extent to which the ICCP capture more genetic diversity by using the number of individual plants per species as a proxy and found that a higher proportion of species were represented by a larger number of accessioned individuals (Fig. 3d). Moreover, the extra resource allocated to capturing more individuals per species is much more apparent for conservation-relevant categories wild-origin and wild-derived, than for garden- or unknown-origin accessions (Fig. 3e). The ICCP seeks to intentionally distribute threatened species across its distributed network of safe sites, as an insurance policy and to empirically evaluate optimum ex situ localities for long-term survival. We therefore asked whether the ICCP threatened species were more globally distributed and so less rare in cultivation, relative to species in the meta-collection as a whole, and found that this was indeed the case (Fig. 3f).

Median survival in the meta-collection is largely immutable

Given the challenging constraints identified in our previous analyses, we then leveraged our longitudinal dataset to understand the persistence of accessions in the meta-collection using Kaplan–Meier survival curves. Here, we adopt a broad definition of survival, to include natural loss as well as deliberate de-accessioning, which together contribute to the transience of living collections. Across the entire meta-collection, we found median survival probability to be 15 years (Fig. 4a), which increased to 20 years for tree collections alone (Fig. 4b). Further segmentation of the data yielded additional insights. For example, we expected to see poorer median survival rates for endemic species due to their more constrained native niche but actually found no difference in median survival between endemic versus non-endemic species, both at 15 years (Fig. 4c). We then evaluated threatened accessions post-onset of digital IUCN Red Lists (1978–2021). Like endemics, threatened species are commonly narrow in native distribution, but in contrast to endemics versus non-endemics, threatened species exhibit a marginally better median survival probability than non-threatened accessions (14 versus 12 years), indicating a small if positive signal of targeted ex situ conservation activity (Fig. 4d). Given that the ex situ conservation of a species over time is contingent on the collective accessions for a given species we further examined the persistence of threatened versus non-threatened species rather than accessions and likewise observed similar improvement in median survival probability for threatened species (16 versus 13 years). Remarkably, we observed the same median survival probability for native and non-native accessions (Fig. 4e), and, contrary to received horticultural wisdom, horticultural cultivars have lower median survival probability (13 years) compared to biological species (16 years), potentially due to selective inbreeding of sport traits unrelated to survival (Fig. 4f). In the context of living collections management, data recording for ‘birth’ (incoming accessions) tends to be more accurate than data recording for ‘death’ (outgoing accessions), which can lag, and so median survival probability values are likely to be upper estimates. But all in all, although certain subcategories of collections such as trees and threatened plants shift the dial to a degree, a median survival probability of ~15 years seems largely immutable across different subsets.