To compile a database of plots in Hawaiian forests (Suppl. material 1), we consulted local experts with extensive experience in forestry, community ecology and botany, as well as experts with knowledge of former and ongoing research projects in Hawai’i, USA. From an initial list of publicly available sources and published studies, we examined each to determine if they met our inclusion criteria. Our inclusion criteria were that each study reports for each plot: i) geographical location, ii) species identity and iii) abundance as the number of individuals of trees, shrubs, and tree ferns. We downloaded raw data or obtained it directly from data owners, which also included individual size either as diameter at 1.3 m (diameter at breast height; DBH) or in size classes. In total, we identified 6 unique studies meeting our inclusion criteria that together comprise 43,590 individuals of 185 species within 530 plots across six islands (Table 1; Zimmerman et al. (2007), Gillespie et al. (2013), Ostertag et al. (2014), Knight & Barton (unpublished), National Park Service’s (NPS) Pacific Island Network (PACN; Ainsworth et al. 2011) and the publicly available data from the U.S. Forest Service (USFS) Forest Inventory & Analysis (FIA) program in Hawaii (US Forest Service 2016). Data were collected between 2003 and 2015.

Studies in the OpenNahele database used different plot sizes and minimum size thresholds (Table 1). Plot sizes ranged from 12.97 m² to 40,000 m² and the median plot size across the database is 1,000 m². Half of the studies used one minimum size threshold and included all individuals above that in their inventory. The other half of the studies used a nested sampling approach, whereby smaller subplots were placed within each plot to assess individuals below the size threshold of the full plot. The minimum size threshold varied across studies from 1 to 2.54 cm DBH. Currently, all studies have only conducted one inventory. Geographic coordinates of all forest plots were converted to the WGS84 coordinate system, a standard coordinate system with a spheroidal reference surface, to facilitate the retrieval of climate, topographical and geological data. Locations of plots in the USFS FIA were fuzzed up to 1.6 km of their exact locations (US Forest Service 2016), but those of other studies were not altered. Geographic coordinates were checked visually.

Most studies inventoried trees, shrubs and tree ferns. While tree ferns do not have true wood, they play an important ecological role in Hawaiian forests (Zimmerman et al. 2008). One study, Knight & Barton, did not record the presence of tree ferns (e.g. Cibotium spp., Sadleria spp.). However, only two plots in that study were located in areas where tree ferns occur and there are many other nearby plots in the database that included tree ferns. As not all studies included lianas, i.e. woody vines, we removed species classified as such by USDA Plants (USDA NRCS 2018).

Data from one study within the OpenNahele database, HIPPNET (Ostertag et al. (2014)), are curated on an ongoing basis and expanded with data from subsequent censuses; interested data users should contact HIPPNET via the CTFS-ForestGEO website.

Abundance

To facilitate aggregation of abundance data across studies that differ in plot size, we calculated abundance of individuals per species on a per-hectare basis:

Abundance per ha = Abundance / Area x 10,000

where abundance is the number of individuals per species and Area is the plot (or sub-plot) area in square metres.

Individual size was converted to centimetres if measured as DBH or classified as greater than or less than 5 cm DBH if individual size was not measured. As data sources used different minimum DBH thresholds, which may influence the number of individuals in a plot in a systematic way, i.e. plots with larger DBH thresholds will have fewer individuals than those with smaller DBH thresholds and, therefore, species diversity estimates, we removed individuals smaller than 5 cm DBH. To further account for variation in the number of individuals due to differences in plot area across the database, we recommend estimating species diversity based on rarefaction curves (e.g. Chao et al. 2014, Chase et al. 2018).

Maximum plant size

We estimated maximum plant size for individual species in two ways: as the 95th percentile of stem diameter of all diameters > 0.1 x maximum observed diameter (D95_0.1) and as the mean diameter of the three largest individuals across the database (D_max3; King et al. 2006). In total, we estimated both measures of maximum plant size for 58 woody plant species that had at least 20 individuals = 5 cm DBH (Suppl. material 2). We compared both measures using Spearman’s correlation coefficient and found that while strongly and positively correlated (rho = 0.89, p-value< 0.001), D_max3 was on average 49% larger than D95_0.1. As D95_0.1 is not sensitive to sample size (King et al. 2006), we recommend using this measure over D_max3, particularly for studies using it as a functional trait in combination with species abundance data.

Taxonomic names were resolved and harmonised with The Plant List v. 1.1 (The Plant List 2013) using the ‘Taxonstand’ package (Cayuela et al. 2017). Family names and orders were also retrieved and used to identify angiosperms and monocots following the Angiosperm Phylogeny Group III (Angiosperm Phylogeny Group 2009). Native status was obtained by consulting the electronic Flora of Hawaii (Wagner et al. 2005). Individuals not identified to the species level were classified as ‘uncertain’, unless the genus was endemic to Hawai’i. As not all introduced species have been naturalised, i.e. they are cultivated in forestry plantations or as ornamentals but have yet to establish self-sustaining populations, we also obtained the cultivation status for introduced species using the Pacific Island Ecosystems at Risk database (US Forest Service 2017). For each individual, accepted species name, family, angiosperm and monocot classification, native status and cultivated status are provided (Suppl. material 1).

The 530 plots in the OpenNahele database are located on all six major islands of the Hawaiian archipelago (Fig. 1). Forested areas are well covered by plots on most islands and include a wide range of habitat types, from tropical dry forests to subalpine shrublands. However, not all islands were sampled with the same intensity (Table 2); 380 plots are located on the largest island, Hawai’i, and the other, smaller islands have between 5 and 59 plots. One potential limitation with this database is that most studies, with the exception of those collected by the USFS FIA, did not locate plots randomly, a fact which may introduce a bias towards forests with relatively low amounts of alien species.

18.90986 and 22.23583 Latitude; -154.8058 and -160.5458 Longitude.

In total, the OpenNahele database contains 185 tree, shrub and tree fern species, of which 61% and 39% are native and alien, respectively, and which represent 16% of the 1,155 woody species that occur across the Hawaiian archipelago (Wagner et al. 2005). The database captures a relatively small proportion of woody species in Hawai’i, possibly because many species may not reach the minimum size limit (5 cm DBH) and that plots were only located in forests and not in ecosystems where woody species occur but are small in stature or are not dominant, e.g. shrublands and grasslands. However, the database has a similar proportion of alien woody species (39%) as found across the entire Hawaiian archipelago (41%; Wagner et al. 2005).

Dominant species

Metrosideros polymorpha is hyperdominant in Hawaiian forests and represents 33% of all individuals greater than 5 cm DBH (Fig. 2; Table 3). Two species of native tree ferns, Cibotium glaucum and C. menziesii, are amongst the five most abundant species in the database. Alien species have heavily invaded Hawaiian forests; fifteen of the 25 most abundant species in the database are alien (Table 3).