We collected data for eight floral traits (Table 1) from 34 published articles and their supplementary materials as well as from two books, reports and dissertations each and from an online database (“References_of_raw_data.pdf”, Dryad repository in folder attachment). For cultivar plants, the traits were sampled mainly in field experiments. For wild flowers, the entities were recorded either on natural, semi-natural habitats or botanical gardens. Most research was done in Europe, especially Poland and England, though some information came from a Northern America database (see “Geographic_Information_raw_data.pdf”, Dryad repository in folder attachment).

For the raw database, some input data of the traits had to be adapted. We calculated the average molar sugar concentration per species from the data of Gilbert (1981) assuming the sugar is pure saccharose. If the values of corolla depth were 0 in the reference of Gilbert (1981) and Becher et al. (2018), the species have open flowers. For species in Becher et al. (2018) which only provide pollen, but no nectar, we set corolla depths to NA. If the nectar volume was 0 in Becher et al. (2018), but other references recorded a nectar volume or nectar sugar content > 0, the values were not transferred into the database.

For the quantitative traits, we gathered minimum, maximum and mean values, if available. With the traits 'pollen', 'nectar volume', 'sugar per flower' and 'flower' or 'inflorescence density', we recorded the flower unit they referred to, i.e. either per single flower or per inflorescence. The reference flower unit is very important for scaling nectar volume, nectar sugar content and pollen to the same flower unit, enabling merging and aggregation of trait data from different sources. Furthermore, the nomenclature of species varied in literature. Therefore, we equalised the species names in our database in column 'species' in our database, but we also included the names used in the original publications to facilitate joins and backtrackings with the data source (column 'species_name_reference' in our database).

Data preparation and multistep workflow

We compiled the FloRes Database so as to include as many species/taxa with a complete set of traits as possible through a multistep workflow in R 4.1.0 (R Core Team 2021) using five consecutive scripts:

1. We converted flowers and inflorescences per square metre as well nectar and pollen per flower or inflorescence to the level of floral units using Formulae 1 and 2 (Script: 1_Inflorescences.R). This step requires the dataset "AgriLand_FlowerDensity_perspecies.csv" of Baude et al. (2015).
2. We converted trait values to the same physical units for each trait and calculated missing trait values from other traits using Equations 3 to 5 and 7 to 9 (Script: 2_units.R).
3. We took the means (except for flowers per square metre where we used the maximum) of multiple trait entries for each species (Script: 3_Aggregate_species.R).
4. We either unified synonymous species names or grouped species on a reasonable taxonomic level (taxon) for the next step to combine and aggregate the plant species. Further, we deleted those with few entries (Script: 4_Selecting_taxa.R). The grouping of species is given in the required auxiliary file “Taxa_to_aggregate.csv”, which can be edited.
5. We calculated the means of the traits of the synonymous species and repeated, now with the more complete dataset, the derivation of traits from other traits using Equations 4 and 6 (Script: 5_Aggregation_selected_taxa.R).

The first and second script were used to convert the data to equal units, whereas scripts three to five were used to aggregate and combine the trait data on the most suitable taxonomic level, preferably on species level (Column 'taxon' in our database). However, we could frequently aggregate only on genus level.

Conversions of units

Throughout the literature, flowers were defined in different units and were given in different units. Therefore, the flower units needed to be transformed to the same definition of floral units (Carvalheiro et al. 2008) per area. This enabled us to scale pollen and nectar availabilty correctly per area.

• Floral units per area

For assessing the quantity of floral resources on habitat scale, the quantity of pollen or nectar (sugar) per standardised area is needed. In the raw data, the floral reference unit, either single flower or inflorescence, sometimes varied between flower-unit density, nectar and pollen data. For simplification the terms raceme, panicle, corymb, globular raceme, umbel or catkin in the database of Baude et al. (2015) were defined by us as inflorescences for the reference floral unit. Therefore, in order to facilitate calculations of nectar or pollen per square metre, we used the entity of the floral units (Carvalheiro et al. 2008) given in Baude et al. (2015). Thus, single flowers could be summed up to inflorescences or inflorescences down to single flowers. Therefore, we transformed abundances of single flowers f_a [m^-2] into abundances of floral units fu_A [m^-2] through division by the number of open flowers per inflorescence f_I using the information from Baude et al. (2015) (Equation 1).

\(fu_A=f_A/f_I \) (1)

• Pollen and nectar per floral unit

Floral resources R_f, i.e. pollen or nectar, were multiplied with the number of open flowers per floral unit f_i (Baude et al. 2015) to obtain the floral resource R_fu per floral unit:

\(R_{fu}=P_f⋅f_i\) (2)

For open flowers per inflorescence of Helianthus annuus, we used data from Minckley et al. (1994), because H. annuus was not recorded in Baude et al. (2015).

For floral units per area, we used the maximum value, not the mean, when there were multiple values per species. Here, we used the maximum density as an approximation for 100% cover of the plant species. This allowed us to scale the floral resource per square metre in a given habitat, when the habitat specific cover percentages of plant species were available. The floral unit density per area needed to be divided through the cover percentage.

• Nectar volume, nectar sugar content and sugar concentration of nectar

Mostly, nectar was measured as secretion of liquid per flower and day [volume flower^-1 d^-1] (e.g. Bosch et al. 1997, Hedtke 2000, Horn 2017, Becher et al. 2018) or as the sugar content per flower and day [mass flower^-1 d^-1] (e.g. Crane et al. 1984, Maurizio and Schaper 1994, Baude et al. 2016, Hicks et al. 2016). We followed this convention and calculated both the volume [ml floral unit ^-1 d^-1] and sugar content per flower [mg floral unit^-1 d^-1]. For the conversion of nectar volume to sugar content and vice versa, we used Equations 3 to 6.

To receive sugar in mass per flower m_z [mg], we multiplied nectar volume V_nec [ml] by density of saccharose ρ_z [1570 mg ml^-1] and sugar concentration c_perc [%]:

\(m_z = \rho_z V_{nec} \frac{c_{perc}}{100}\) (3)

When only molar concentration of sugar c_mol [mol l^-1] was given, we used the following equation with molar mass of saccharose M_z [342.3 mg mol^-1]:

\(m_z = V_{nec} M_z \frac{c_{mol}}{1000}\) (4)

To calculate the nectar volume V_nec [ml] from sugar content per floral unit m_z [mg], we used:

\(V_{nec} = \frac{m_z} {\rho_z⋅c_{perc}} 100\) (5)

When c_perc was not given, we used c_mol to calculate V_nec:

\(V_{nec} = \frac{m_z} {M_z c_{mol}} 1000\) (6)

For our application, we needed the sugar concentration in molar concentration, so we transformed c_perc to molar concentration as:

\(c_{mol} = c_{perc} \frac{\rho_z} {M_z} 10\) (7)

• Mass of pollen per floral unit and pollen protein content

For pollen, the physical units differed due to extraction methods. Mostly, the mass of pollen was given and the values only needed to be scaled to mg, if given in g or µg. However, sometimes it was given as estimated volume of pollen grains (Hicks et al. 2016). Therefore, we used:

\(m_p=V_p * \rho_p\) (8)

to calculate pollen mass m_p [mg] from the pollen volume V_p [ml] and the pollen density ρ_p [mg ml^-1]. Since for most species, the density of fresh pollen ρ_p is not known, we used:

\(\rho_p = \rho_{prot} P_{prot} + \frac{(1 - P_{prot})}{3} (\rho_{starch} + \rho_{fat} + \rho_{water})\) (9)

with mean densities of protein ρ_prot [1300 mg ml^-1] (Chick and Martin 1913), of starch ρ_starch [1440 mg ml^-1] (Marousis and Saravacos 1990), fat ρ_fat [900 mg ml²] (Deutsches Institut für Normung e.V. 2002) and water ρ_water [1000 mg ml^-1] and with the proportion of protein of the pollen P_prot[-]. When P_prot was not given for a plant species, we estimated it through mean protein content of the genus, as the protein contents are relatively similar amongst the species of a genus (Roulston et al. 2000). When the protein amount could not be estimated for a genus, it was estimated as the mean of all species in the database.

Aggregation of data, replacement of synonyms and FloRes Database

After equalising floral, physical and chemical units (script: 2_units.R), we aggregated the traits using the mean of multiple entries per taxon, except for the case of floral units per area, where we used the maximum value to receive an approximate density of floral units at 100% coverage of the species (script: 3_Aggregate_species.R). Subsequently, we checked the species for completeness of traits and grouped closely-related taxa with incomplete, but complementary, trait information in a table (Taxa_to_aggregate.csv) for further aggregation on genus level or a reasonable higher taxonomic level. We used this table to add how plant species should be automatically aggregated in the script 4_Selecting_taxa.R. Moreover, we used this step to aggregate the synonymous names of a species with their common name or on a higher taxonomic level. Finally, we aggregated the traits a second time by the selected taxon (species, genus or higher level). In cases where values for molar sugar concentration were still lacking after Step 5, we inserted the average value of 40% of sugar concentration as an estimate for wildflowers, as given in Westrich (2018).

After the final aggregation, we got three different output tables for the FloRes Database. “5_FloRes_raw” contains the mean values of all taxa for which at least some trait data were available. “5_FloRes_complete_trait” is the dataset of taxa without any gaps. “5_Selected_taxa_no_corolla” contains taxa where all traits, except for corolla depths, were complete. Those datasets can finally be used to calculate the amount of nectar and pollen of habitats within any defined time period, given that the plant species of the habitats are included in the database.

The database is a collection of data from the Northern Hemisphere, focused on Central Europe. The details about the geographical information of the raw database references are listed in https://datadryad.org/stash/share/pYjuf_kRaA0N9Lw25svZa_rnQ_mENIIyQAC2rkXicEI.

The FloRes Database: A floral resources trait database for pollinator habitat-assessment generated by a multistep workflow

https://doi.org/10.5061/dryad.djh9w0w29

2