Implications for conservation assessment from flux in the botanical record over 20 years in southwest Ghana

Abstract At best, conservation decisions can only be made using the data available at the time. For plants and especially in the tropics, natural history collections remain the best available baseline information upon which to base conservation assessments, in spite of well‐documented limitations in their taxonomic, geographic, and temporal coverage. We explore the extent to which changes to the plant biological record over 20 years have changed our conception of the conservation importance of 931 plant taxa, and 114 vegetation samples, recorded in forest reserves of the southwest Ghana biodiversity hotspot. 36% of species‐level assessments changed as a result of new distribution data. 12% of species accepted in 2016 had no assessment in 1996: of those, 20% are new species publications, 60% are new records for SW Ghana, and 20% are taxonomic resolutions. Apparent species ranges have increased over time as new records are made, but new species publications are overwhelmingly of globally rare species, keeping the balance of perceived rarity in the flora constant over 20 years. Thus, in spite of considerable flux at the species record level, range size rarity scores calculated for 114 vegetation samples of the reserves in 1996 and 2016 are highly correlated with each other: r(112) = 0.84, p < .0005, and showed no difference in mean score over 20 years: paired t(113) = −0.482, p = .631. This consistency in results at the area level allows for worthwhile conservation priority setting over time, and we argue is the better course of action than taking no action at all.


| INTRODUC TI ON
At best, conservation decisions can only be made using the data available at the time. Short of surveying a whole region to make decisions locally, the best we can hope to do is to use all available data and to characterize how our current data may be biasing our perspective. For plants, the most commonly used baseline information upon which conservation assessments are based is the geographic distribution of species in terms of area of occupancy (AOO) or extent of occupancy (EOO) as revealed by herbarium specimens . While there have been significant efforts to digitize and curate herbarium records (Dauby et al., 2016;Enquist et al., 2016), the expense of collecting and storing specimens long-term has limited the plant biological record in terms of its taxonomic, geographic, and temporal coverage and resolution (Meyer et al., 2016;Roberts et al., 2016;Schmidt-Lebuhn et al., 2013;Stropp et al., 2016).
The principle problem with an incomplete baseline description of plant species' distributions for conservation is that newly accessioned specimens (or new field observations) will tend to produce increases in EOO and AOO .
If our evidence threshold for conservation assessment is too stringent, we fail to define a baseline against which to monitor change for many species. If our evidence threshold is set too low, we risk producing misleading or labile conservation assessments inappropriate for a legal framework. Protecting "undeserving" species has economic costs, while failing to protect "deserving" species results in local or global extinctions. At the moment, our evidence threshold for conservation assessment via the IUCN Red List is set very high for plants. Authors need "good evidence to establish that a species is not undercollected" in order to publish an IUCN Red List assessment for a species with <5 records , while 15 georeferenced specimens per species are needed to produce Red List range estimates consistent with estimates based on all known specimens (Rivers et al., 2011). Fifteen distinct georeferenced specimens are a high bar for most plant species. It takes around 70 years from the collection of the first specimen to reach 15 specimens (Goodwin et al., 2020). 36.5% of the world's plant species (158,535 species) are represented by five herbarium or field observations or fewer, with 28.3% (123,149 species) having three observations or fewer, and 13.6% to 11.2% species with just one observation (Enquist et al., 2019). The result of this high burden of evidence is that Red Listing progress for plants has been limited, at around 8.5% of accepted vascular plant species names, excluding those in need of updating (26,720 of 316,143). As much as 39% of plant species selected randomly for inclusion in the Sampled Red List Index for Plants could not be evaluated against the Red List criteria, even as Data Deficient .
For the 5 years preceding 2019, substantially fewer species were assessed for the Red List each year than were published the same year, resulting in negative progress in real terms after the historical synonymy rate of 0.34% is applied.
In Ghana, the Star system has been used since 1996 to prioritize plant species for conservation (Hawthorne, 1996). In the Star system, species are assigned to one of four categories of global range size rarity by reference to their area of occupancy (AOO) measured globally with a fixed cell size of 1 degree square (or 100 × 100 km outside the tropics). Black Star species are the most globally rare species, with a mean global AOO of 2.7 degree squares. Gold Star species have restricted ranges of 8 degree squares on average, Blue Star species have a three times larger range of 24 degree squares, while Green Star species are the most widespread. One of the tenets of Star rating is that all species should be rated using the best information available at the time, even if this is perhaps incomplete, with ratings being updated in light of new information if necessary . There is also no requirement to show a decline in population size or distribution, as there is for the Red List, which reduces the burden of proof further. All plant species in tropical Africa have a published Star rating . In Ghana, among other countries, area-based conservation action has also been taken on the basis of the Star ratings. Each Star category carries a weight in inverse proportion to its rarity. The weights of the species present in a vegetation sample are averaged to give a bioquality score for the vegetation overall, reflecting the concentration of globally rare plants within the vegetation. In 1999, the Ghanaian Government established the High Forest Biodiversity Conservation Project (HFBCP) funded by the World Bank, which aimed to conserve biodiversity in these forests through the establishment and protection of Globally Significant Biodiversity Areas (GSBAs). These reserves were identified and protected on the basis of their high bioquality scores. This approach led to the establishment of 29 forest reserves amounting to c. 2300 km 2 of forest reserves or 13% of the total forest network and their ex-  (Brooks et al., 2006;IUCN, 2016a;Myers et al., 2000). There has been some work addressing how uncertainty in our knowledge of plant species' taxonomy and distribution manifests in lability of species' IUCN ratings over time (Goodwin et al., 2020;Lughadha et al., 2019;Rivers et al., 2011). However, how incomplete knowledge at the species level manifests in conservation assessments of areas is an important and underexplored question (Daru et al., 2020;Murray-Smith et al., 2009;Nelson et al., 1990).
Here, we describe changes in the nomenclature and documented distributions of the vascular plant species of the southwest Ghana forest reserves over 20 years (1996 to 2016). We explore the consequence of these changes for species-level conservation assessments (Star ratings) and area-based conservation assessments (bioquality scores). We argue that in spite of considerable flux in the botanical record, area-based conservation assessments in particular have remained remarkably stable over 20 years and have led to positive outcomes for conservation in Ghana. By contrast, forests outside of reserves experienced significant and growing threats from logging, fuelwood collection, agriculture, mining, and climate change (Aleman et al., 2018;Edwards et al., 2014;McClean et al., 2006).

| Study area
Upper Guinea is a phytogeographical name for the forest zone of west Africa, running from Sierra Leone in the west to the Dahomey Gap (Ghana) in the east, from the coast up to 350 km inland (Marshall et al., 2021;White, 1979). This area is included within the Western African Forests ecoregion, which has been recognized as a biodiversity hotspot (Myers et al., 2000;Olson et al., 2001). The forest reserves of SW Ghana, where this study is situated, have also been recognized as Key Biodiversity Areas (KBAs) (Key Biodiversity Areas Partnership, 2022). Southwest Ghana is regarded as a hotspot of plant endemism within west Africa, being home to a high concentration of globally rare species (Bongers et al., 2004;Marshall et al., 2016Marshall et al., , 2022. The typical flora of the study area is lowland evergreen rainforest, with variations in species composition driven by gradients in rainfall, disturbance, and local topology, with altitude and historical climatic stability important at broader geographical scales (Marshall et al., 2022). Across the sampled sites, annual precipitation averages 1910 mm/year, while mean temperatures average 26.5°C (Hijmans et al., 2005). Altitude above sea level has a minimum of 40 m, a maximum of 181 m, and a mean of 97 m. The forest reserves of southwest Ghana are subject to anthropogenic disturbance, for example, accommodating logging concessions.

| Field methods
Twenty-seven samples were enumerated in 2015 for this publication and were compiled with 87 botanic samples from the region from different surveys ( Table 1)

| Species
We use Star ratings as our species-level conservation assessment Marshall et al., 2016). Species are assigned to one of four categories of range size rarity, measured by reference to their area of occupancy (AOO) measured globally with a fixed cell size of 1 degree square (or 100 × 100 km outside the tropics Star ratings for 1403 species of the forest zone of Ghana were published in 1996, using the best available distribution data available at the time (Hawthorne, 1996). Nine of those records were then redundant, as they had been sunk into names already listed (1 Black, 2 Gold, 2 Blue, and 4 Green).
Homotypic synonyms with a merely cosmetic name change and no associated range change are treated for this analysis as if they remained unchanged, that is, they are not included in the counts of F I G U R E 1 Map of 114 sample locations in the southwest Ghana study area, colored by GHI calculated using the Stars published in 2016 and including all species found. Reserves are named.
changes or additions. Reddish Stars (S, R, P) and savanna (X) Stars published in 1996 were excluded from the dataset and calculations, as these are no longer in use.
A species' Star rating may change for two reasons: (i) new distribution data alters our understanding of the species' range; (ii) the species experiences a genuine increase or decrease in range size.
Star ratings may be added when (1) new species are published (and found in southwest Ghana), including the splitting of taxa; (2) previously published species are newly recorded in (southwest) Ghana; (3) "known unknowns" are resolved, that is, species which were considered to be of uncertain Star in 1996 are resolved to a Star in 2016 thanks to new sources of information. This flux is visualized with a chord diagram implemented with R package circlize (Gu et al., 2014; R Core Team, 2020).

| Areas
We use the Genetic Heat Index (GHI) as our area-level conservation rating (Hawthorne, 1996;, the standard index of bioquality published in 1996 and used to designate Ghana's GSBAs. GHI is a continuous metric representing the weighted global ranges (endemism) of the species present in a plant community. The calculation is similar to range size rarity, except that species' ranges are measured globally at one degree square resolution rather than within the study area at any resolution, species ranges are categorized into four groups, called Stars, rather than being treated continuously, and bioquality includes no measure of species richness, as is sometimes the case with range size rarity metrics.
GHI is calculated as the proportion of species belonging to each Star rating within a sample, where each species is inversely weighted by the mean range size of its Star (BK = 27, GD = 9, BU = 3, GN = 0). The GHI for each sample was calculated from the species present, using the following formula, where NBK, NGD, NBU, and NGN are the number of Black, Gold, Blue, and Green Star species in a sample, and WBK, WGD, and WBU are the respective weights.

| Stability of species-level conservation assessments
Changes and additions to published Star ratings over 20 years are explored for the forest zone of southwest Ghana (   Downgrading of species from the rare categories to more widespread categories is the result of increases in the documented range of species, due to more distribution data records being collected and becoming available online and for analysis. An inestimable proportion of these records will represent genuine range extensions for species, although the majority are very likely to be simply the result of better recording of established ranges. Overwhelmingly, the upgraded species are species which moved from Green to Blue Star: These are species which appeared to be widespread from their range description in FWTA (e.g., "to Lower Guinea"), but on consideration of their modern dot map distribution data have subsequently been shown to be sparse within that extent of occurrence.

| Additions
Of the 931 species now considered to be present in samples of southwest Ghana, 109 species (12%) had no Star rating in 1996 (Table 4).
New additions to the flora are disproportionately rare species: A chi-square test of association shows that many more species additions proved to be Black or Gold Star species than would be expected given previously rated species, and many fewer than expected proved to be Green Star (

| Conservation priority species
Changes to the list of species considered to be of the highest conservation priority are detailed in Table 5.

| Inferred genuine changes in conservation status
The  refound: These are potentially of concern, but given the lower sampling effort and the local rarity of these species we do not believe that

| DISCUSS ION
This study showed that there has been considerable flux in our conceptions of taxonomy and distributions for plant species within southwest Ghana over 20 years, an area that is well studied relative to other tropical endemism hotspots. 36% of species changed global AOO category (Star category), and 12% of species occurrences in our dataset have been contributed since 1996. Species tended to appear more widespread, rather than more restricted, over the 20-year re-evaluation period (20% vs 16%). This effect is most marked for the globally rarest species, with two thirds of local endemics moving to a more widespread AOO category after 20 years and just one third remaining in the same category. However, new species discoveries and new records for the region were disproportionately of globally rare species, so that the total number of local endemics known in southwest Ghana was maintained over the 20 years. This explains Note: Species denoted with asterisk * are species large enough trees to be of potential interest for timber, but are nevertheless still protected from logging as they were designated Black Star in 1996. All are also Red Listed threatened species, EN, Endangered, VU, Vulnerable, and we do not recommend that their protection from logging is removed. This lability in documented ranges and conservation assessments is probably applicable to most groups of plants in the tropics.

TA
In neotropical Myrcia, for example, around 33% of species changed their apparent EOO-based conservation status as the result of 10 years of specimen collection, databasing, and taxonomic changes . Similar to our findings, transitions to smaller EOO categories outnumbered transitions to larger EOO categories (8% vs 3%, respectively), and species previously deemed data deficient transitioned to threatened status more often than to not threatened (10% vs 7%, respectively; Nic Lughadha et al., 2019).
For Madagascan orchids, species described more recently have smaller ranges and occupancies, fewer specimens and greater perceived extinction risk status (Roberts et al., 2016). In the Cape Floristic Region, narrow-range taxa have constituted a significantly greater proportion of species discoveries since 1950 (Treurnicht et al., 2017). The almost universal lability of taxonomic nomenclature has been perceived as awkward for conservation (Garnett & Christidis, 2017). Rather than indulge calls to "finalise" taxon names (Thomson et al., 2018), accepting a degree of lability in conservation assessments would be an alternative and more accurately reflect our evolving knowledge of the natural world.
Although some relatively common species were previously considered rare in our study, as a result of undersampling, no rare species were misclassified as common, so all the species that appeared worthy of attention at the time would have received it. Many rare species had been overlooked or were not yet described, and these were disproportionately found in original hotspot locations, despite equal resampling effort in cold and hot spots. The precautionary principle is often touted in conservation science (IUCN, 2016b), but it less often followed. When specimens represent the best available evidence for particular species, their use as a basis for extinction risk assessment is appropriate, necessary, and urgent . This case study supports the precautionary principle: The correct areas were given conservation funds despite undercol- threatened plant species, we support calls to assess plant species using the evidence available and characterize how our current data may be biasing our perspective, but ultimately to accept some lability in those assessments.
We recognize that the goal of the Red List is to assess extinction risk in a comparable way across eukaryotes, rather than to assess a taxon's range (Collen et al., 2016). The relationship between the risk of extinction and global range (endemism) is not absolute, though narrow AOO or EOO is an all but essential prerequisite for a plant's assessment as threatened, and many restricted range species are threatened (Robbirt et al., 2006). Tropical plant species with narrow ranges declared extinct are least likely to be discovered extant (Humphreys et al., 2019). We suggest that species known from <5 specimens should not be considered as unassessable, or Data Deficient. The precautionary principle, which is already enshrined in the IUCN guidance, should be employed in practice. Such species could be assessed by their AOO, along with number of locations, and appropriate proxies or evidence for a decline in habitat area, extent or quality which could be relatively general in the case of plants.
Our study suggests that such species may indeed prove to be more widespread (and less threatened) over time than they initially appeared.
But in the meantime, their recognition and presence shines a light on an area which is poorly known and quite likely to harbor new records or species new to science, which has the added benefit of being afforded conservation protection, perhaps also helping to reduce the historical 70 years required to collect 15 geolocated specimen localities (Goodwin et al., 2020).
Herbaria are principally a taxonomic repository, retaining type specimens and significant range extensions such as new country records. Vouchered specimens remain essential to provide verifiable identifications for work in ecology, phylogenetics, taxonomy, conservation, and pharmacology (Funk et al., 2018). A new sort of herbarium specializing in, or at least accepting, recollections of species from known localities, specimens located within the known EOO, sterile vouchers, and specimens which describe the species of a locality in depth would be useful for conservation assessment and reassessment. Such specimens need not be curated to such a high specification as traditional specimens or even be kept at all, once photographed with a small amount of leaf material retained, for example, for DNA extraction. Specimens could be linked via QR codes or similar to online platforms such as iNaturalist, providing supplementary access to living plant photographs, identification history, location maps, and potentially additional geolocated records from the area, without increasing physical storage requirements (Heberling & Isaac, 2018).
We further suggest a greater emphasis is placed on area-based conservation assessment than species-level assessment (Plumptre et al., 2019). Habitat assessment is an extension from species-level assessment, an approach adopted in the Star and bioquality system, as well as Key Biodiversity Areas and Tropical Important Plant Areas Key Biodiversity Area assessment in Uganda showed that most of the remaining natural habitat was important for the conservation of globally and nationally threatened species and threatened habitat (Plumptre et al., 2019). Emphasizing area-based assessments, such as range size rarity or weighted endemism, also reduces the importance of "fixing taxonomy" at the species level (Thomson et al., 2018).
Our study shows that locality assessments, which make use of the summed (or averaged) signal from many species, were more stable over time that the assessment of any one individual species. By acting on the signal from the apparently rare species, some of which later turned out to more widespread, conservation priority areas were nevertheless identified and protected. It is surely not wise to imagine that we can ever finalize our estimates of AOO, EOO, Red List status, or nomenclature. Fortunately, the consequences of not doing so may be less severe than we imagine.