A dataset of the flowering plants (Angiospermae) in urban green areas in five European cities

This article summarizes the data of a survey of flowering plants in 80 sites in five European cities and urban agglomerations (Antwerp, Belgium; greater Paris, France; Poznan, Poland; Tartu, Estonia; and Zurich, Switzerland). Sampling sites were selected based on a double orthogonal gradient of size and connectivity and were urban green areas (e.g. parks, cemeteries). To characterize the flowering plants, two sampling methodologies were applied between April and July 2018. First, a floristic inventory of the occurrence of all flowering plants in the five cities. Second, flower counts in sampling plots of standardized size (1 m2) only in Zurich. We sampled 2146 plant species (contained in 824 genera and 137 families) and across the five cities. For each plant species, we provide its origin status (i.e. whether the plants are native from Europe or not) and 11 functional traits potentially important for plant-pollinator interactions. For each study site, we provide the number of species, genera, and families recorded, the Shannon diversity as well as the proportion of exotic species, herbs, shrubs and trees. In addition, we provide information on the patch size, connectivity, and urban intensity, using four remote sensing-based proxies measured at 100- and 800-m radii.

We sampled 2146 plant species (contained in 824 genera and 137 families) and across the five cities. For each plant species, we provide its origin status (i.e. whether the plants are native from Europe or not) and 11 functional traits potentially important for plant-pollinator interactions. For each study site, we provide the number of species, genera, and families recorded, the Shannon diversity as well as the proportion of exotic species, herbs, shrubs and trees. In addition, we provide information on the patch size, connectivity, and urban intensity, using four remote sensing-based proxies measured at 100-and 800-m radii.  Table   Subject Ecology, Nature, and Landscape Conservation. Specific subject area Urban ecology Type of data Table  Fig. How data were acquired Floristic inventories and standardized floral counts. Satellite data. Data format Raw and aggregated Parameters for data collection Sites were selected from the European Urban Atlas, using the features mapped as green areas. Sites were chosen following an orthogonal gradient of patch size and connectivity inferred with the proximity index. We selected 32 sites in Zurich, Switzerland, and 12 sites in each of the remaining four cities (i.e. Antwerp, Paris, Poznan and Tartu).

Description of data collection
We applied two sampling methodologies inside of buffers of 100 m radius: 1) a floristic inventory of the occurrence of all flowering plants of potential interest for pollinators performed in the five cities, and 2) flower counts in sampling plots of standardized size (1 m 2 ) done only in Zurich. Sites were visited on three occasions between April and July 2018.

Value of the Data
• The dataset describes the diversity, occurrence, and floral counts of a large number of flowering plant families sampled in a standardized way in different types of public and private green areas in European cities, and with a high taxonomic resolution.
• The data contribute characterizing European urban floras, derive taxonomic, phylogenetic and trait diversity patterns, and perform comparative studies among different cities, different types of urban green areas and in fragmentation studies. • The data can be used to characterize the available food resources of other trophic levels, particularly pollinators, and species interactions. • The data on floral counts can be combined with metrics on nectar and pollen content to obtain estimates of resources quality (e.g. as done in [1] ) • The methodology for collecting the data can be applied in further studies aiming to characterize plant resources in one or more urban ecosystems in a standardized way.

Data Description
The paper presents the data of a plant survey in urban green areas from five European cities and urban agglomerations (Antwerp, Belgium; greater Paris, France; Poznan, Poland; and Zurich, Switzerland). 80 sites were selected (32 in Zurich and 12 in each of the remaining four cities, see Fig. 1 ) according to an orthogonal gradient of patch size and connectivity (see Section 2.2 ), representing common public urban green areas such as parks, cemeteries and gardens. To characterize the flowering plants, we sampled plants during four (for Zurich) and three (for Antwerp, Paris, Poznan and Tartu) sampling periods during the year 2018. The sampling was performed in (1) end of April (only for Zurich), (2) end of May, (2) end of June and (3) end of July. The sampling consisted in two methodologies. First, a floristic inventory of the occurrence of all flowering plants inside buffers of 100 m radius (see Fig. 1 ) in the study sites of the five cities. Second, flower counts of defined floral units ( Table 1 ) in sampling plots of standardized size (1 m 2 ) distributed inside buffers of 100 m radius (see Fig. 1 ) done only in Zurich. The 100 m radius buffer was defined from existing installed trap-nests place to sample cavity-nesting bees and wasps ( Fig. 1 ).
For each of the 2146 plant species recorded we show in what cities it was recorded ( Supplementary material, Table A1 ). Furthermore, we provide information on 11 traits of potential interest to study plant-pollinator interactions ( Table 2 ) that are the flowering duration, flowering start, growth form, inflorescence type, plant height, floral rewards in the form of nectar, oil and pollen, structural blossom class and floral symmetry based on bibliographic information. Additionally, we documented the origin status of all the sampled plant species, that is, whether or not they are native from Europe. We computed the species, genera, and family richness for each site ( Table 3 and Fig. 2 ) and the composition of plant families of the species sampled in each city ( Fig. 3 ). Moreover, we computed the proportion of exotic species, as well as the proportion of trees, shrubs, and herbs for each site and city ( Table 3 and Fig. 4 ). In addition, we show the frequency distribution of floral counts ( Fig. 5 ) and the composition of plant genera in the flower abundance ( Fig. 6 ) in the city of Zurich.
We provide information on the study site features including the city, their size, connectivity, and urban intensity inferred using a set of remote sensing-based proxies on soil, grey infrastructure, and vegetation, including the Second Brightness Index (BI2), the Color Index (CI), the Urban Index (UI), and the Normalized Difference Vegetation Index (NDVI) within a 100 and 800 m buffer centered in the centroid of the urban green area ( Table 4 ). The data are part of the interdisciplinary research project BioVeins investigating different aspects of urban biodiversity and ecosystem services in urban green areas in European cities ( https://www.biodiversa.org/1012 ). The data can be linked to other taxonomic groups such as nocturnal insects and bats [2] , sampled in the same study locations and during the same period. The raw data are available from the repository Envidat [3] with the DOI doi:10.16904/envidat.210.

Table 1
Definition of the flower units and calculation of the floral abundance. For each floral unit type, we show the plant taxa included and how the floral abundance was calculated. For specific floral unit types (i.e. capitula in Dipsacoidae, compound cymes, corymb, panicles, racemes and umbels) we estimated the number of flowers per floral unit by counting all the flowers in seven floral units and computing the mean. Ditto = the same again.

Floral unit definition
Plant taxa
Herb Shrub Tree Climber [6 , 9-11] Inflorescence type Determines whether the blossom is a single flower or an inflorescence.

Site selection
We selected patches among urban green areas mapped and defined in the European Urban Atlas [see 20 ], which includes mostly public urban green areas in the form of parks, cemeteries, and ruderal patches. We used an orthogonal gradient of patch size (area in m2) and connectivity. Connectivity was calculated using the Proximity Index (PI) which considers the area and the distance to all nearby patches with a favorable habitat, within a given search radius (in our case 50 0 0 m), and is defined as: Table 3 Summary statistics of the plants recorded. For each of the 80 study sites in the five cities, we provide the number of species ( N species ), genera ( N genera ), families ( N families ), the value of the Shannon diversity index ( H' ), and the proportion of herbs ( P herbs ), shrubs ( P shrubs ), trees ( P trees ), and exotic species ( P exotic ). H' was calculated using the frequency of each plant species, obtained as the number of cells from the total 16 where the plant was found. The data are based on floristic inventories in the study sites. The data are plotted in Fig. 1 -2 . Site codes represent the study sites shown in Fig. 1 . Note that the statistic does not include species in the families Cyperaceae, Juncaceae and Poaceae. The coordinates of the sites are provided in Table 4   Where a ijs is the area (m 2 ) of a patch ijs within specified neighbourhood (m) of a patch ij , and h 2 i js is the distance (m) between the patch ijs , based on patch edge-to-edge distance. Thus, the PI measures the degree of patch isolation, with highest values given to less isolated patches. We considered as favourable habitat all patches with high probability of having trees (besides urban green areas, also forest and low density urban, with less than 30% impervious surface, see [20] ). The search radius was set to 5 km from each focal patch, the maximum possible with the available cartography. In fact, lower buffer values (from 500 m onwards) did not greatly change the PI values, because the distances are squared, thus greatly limiting the impact of patches beyond a certain distance. To select patches using the orthogonal design, all possible patches were classified in six size classes and six classes of the PI (36 possible combinations). Within these combinations, patches were selected randomly (random stratified sampling design). Due to resource limitations, we only used 1/3 of the possible combinations in Antwerp, Paris, Poznan, and Tartu (maximizing the gradient) and the full range of combinations in Zurich (32 combinations, the other combinations were not available in the city). This resulted in the final selection of 80 sites ( Fig. 1 Table 4 Site features for each study site based on remote sensing data. For each of the 80 study sites in the five cities, we provide the coordinates where the trap-nest was located, the proximity index (Prox), and patch area (Area) used to select the study sites, and the values of the Second Brightness Index (BI2), Color Index (CI), Urban Index (UI), and Normalized Difference Vegetation Index (NDVI) at 10 0-and 80 0-meter radii.

City
Site   [18] and boxplots representing the Shannon diversity (A) and the number of families (B), genera (C), and species (D) recorded in the study sites in Antwerp, Paris, Poznan, Tartu, and Zurich. Each point in a city represents a measurement in one of the sampling sites (12 in Antwerp, Paris, Poznan and Tartu and 32 in Zurich) and in one of the sampling periods (four periods for Zurich and three periods for the remaining four cities). Note that for Paris and Zurich there are two points with larger richness, which corresponds to the study sites in the botanical gardens of Paris (Jardin des Plantes, National Museum of Natural History) and Zurich (Zurich Botanical Garden). Note that the families Cyperaceae, Juncaceae and Poaceae were not included in the sampling.

Remote sensing indices
Urban intensity has been inferred using remote sensing indices on soil, impervious surfaces and vegetation. Particularly, we used the BI2, CI, UI, and NDVI. The BI2 results from the following equation: Where ρRED, ρGREEN and ρNIR are the responses in red, green, and near-infrared bands, respectively. This index is sensitive to the brightness of soils, which in turn is influenced by soil moisture, presence of salts and organic matter content on the soil surface. Thus, brightness values greater than 0.3 are an indicator of soil problems with less decomposed organic materials, which can be reflected in a lower development of trees. In turn, low values of brightness are associated with soils with high moisture content and decomposed organic materials, favoring the growth of tree plants.
The Colour Index (CI) was introduced by Pouget et al. [21] and results from the following equation: Where ρRED and ρGREEN are the responses in the red and green bands, respectively. Although this index was developed to differentiate various types of soils in arid environments, it can help to compute better vegetation indices for incomplete canopies. In most cases, the CI provides complementary information with the BI2 and the NDVI, allowing to differentiate plants and soil more effectively, especially in study areas with less than 10% vegetation [21] . Typically, low CI values have been shown to be correlated with the presence of a high concentration of carbonates or sulfates, nutrients that can serve as fertilizers for plant growth. Meanwhile, higher values have been correlated with crusty and sandy soils and with a low content of organic matter. Thus, this index seems to be a good indicator of soil degradation.
The Urban Index (UI) was developed by Kawamura et al. [22] to effectively detect the structural details of urban cores. The UI was calculated using the following equation: Where ρMIR2 and ρNIR are the responses in the second short wave and near-infrared bands, respectively. Thus, it is a good index for detecting built and non-built areas and can also be used to identify building densities. The built-up area tends to have UI values greater than 0, while negative values close to −1 tend to be green areas.
Finally, the NDVI was developed by Tucker [23] and is the one of the most common indices widely applied for monitoring vegetation dynamics. This index results from the following equation: Where ρNIR and ρRED are the responses in near infrared and red bands, respectively. This index indicates the photosynthetic capacity, or the energy absorbed by plant canopies, hence, the amount of healthy vegetation. Thus, higher NDVI values indicate a higher density of green vegetation. Specifically, in urban environments, NDVI values greater than 0.5 correspond to vigorous green areas, while NDVI values between 0.2 to 0.5 indicate moisture-stressed vegetation, such as natural meadows. NDVI values near zero and decreasing negative values indicate non-vegetated features, such as artificial and barren surfaces, water bodies, snow, and clouds.
These four indices can be used to charecterize the existing vegetation and urban infrastructure in the vicinity of sampling sites. Remotely sensed data can be used to improve research on biodiversity and ecosystem services, being a valuable tool to support more sustainable urban planning and management.

Floristic inventories
Between April and July 2018, we sampled all available plants of potential interest for pollinators (i.e. we excluded the families Cyperaceae, Juncaceae and Poaceae) in a buffer of 100 m radius around each trap nest on green areas both public and private within the defined radius in three sampling periods. Each buffer was divided in 16 cells (see Fig. 1 ). In each buffer, we documented all the plant species found, in order to obtain an estimate of both plant richness and frequency (as the number of cells inside a buffer each species was found). To identify plant species, we used identification guides for European [14 , 15 , 24] and Swiss flora [25] , as well as specialized guides for ornamental plants [13 , 16 , 26] and previous species inventories, e.g. [9] in Zurich. The  Table 1 for the definitions) in all the quadrats for a given site and sampling period, giving a total N of 128 (32 sites x 4 sampling periods). The dashed vertical line represents the median floral abundance and the straight vertical line the mean floral abundance recorded. total duration of each sampling in a site was restricted to about 2.5 h to standardize sampling effort. Note that the late winter and early spring flowers were missed (e.g. Crocus spp., Galanthus spp.). The species, genus, and family richness of each sampling site are given in Table 3 . The list of all taxa and the number of observations per taxon are given in the Supplementary material, Table A1 .

Floral counts on standardized plots
We calculated the floral abundance in a site and sampling period by using 1 m 2 quadrats randomly distributed inside the 100 m buffer. The number of quadrats was determined according to the amount of green areas in each buffer, with a minimum of seven quadrats, when less than 20% of the buffer was covered by green areas, and a maximum of 15 quadrats, when more than 90%. To obtain the floral abundance, we first defined a set of floral units on where we classified the different plant species (see Table 1 ). The floral abundance of each floral unit type was calculated in the following way. For single flowers, the floral abundance was obtained by summing all the individual flowers ( Table 1 ). For single capitula (in Dipsacoideae species), single compound cymes, single corymbs, single panicles, single racemes, single umbels, we took seven different floral units, counted all the flowers and computed a mean number of flowers per floral unit. The mean number of flowers per floral unit was calculated separately for each site and sampling period. The floral abundance was then obtained by multiplying the number of floral units and the mean number of flowers per floral unit (see Table 1 ). Finally, for single capitula (in Asteraceae), single catkins, single corymbs or cymes in the Euphorbia genus, single dense clusters (including only Sangisorba spp.) and single spikes (including the genus Plantago spp. and Tamarix spp.) the floral abundance was computed by summing only the number of floral units, that is, we did not estimate a mean number of flowers per floral unit.

Taxonomic treatment
Taxonomy assignment largely followed the criteria of Checklist of the National Data and Information Centre of the Swiss Flora [27] , together with The World Flora Online database [28] , and other resources, e.g. RHS Dictionary of Gardening [16] . Varieties, taxa within species complexes, and cultivars were mostly grouped into aggregates (e.g. Taraxacum officinale aggr.) or left at the genus level (e.g. Leucanthemum sp.) without further distinction.

Traits
We aimed to select important determinants of plant-pollinator interactions. We developed a data set of 11 traits (see Table 2 ) for 2313 plant species. We used 11 functional traits including start and duration of the flowering period, growth type, inflorescence type, pollination mode, blossom class, symmetry, plant height and the presence of rewards in the form of pollen, nectar and oils. Additionally, we included the origin of the plant species, which are no functional traits per se. For functional traits, we used as main sources the TRY plant trait database [10] , the national data, and information centre for the Swiss flora [6] , the Bundesamt für Naturschutz (BfN) field technicians Piotr Kazimirski, Laure-Anne Frank and Emeline Klimczak. We thank Simone Fontana, Jordi Bosch, and Laura Roquer for their input developing the methodology. We thank the municipalities of Antwerp, Greater Paris, Poznan, Tartu, and Zurich for their engagement on the project BioVEINS.

Supplementary Materials
Supplementary material associated with this article can be found in the online version at doi: 10.1016/j.dib.2021.107243 .