Small area and low connectivity constrain the diversity of plant life strategies in temporary ponds

(i) To determine whether area and connectivity of temporary ponds can predict plant species diversity, and the diversity and abundance of different plant life histories; (ii) To explore whether pond connectivity with the river prior to river regulation predicts better plant diversity patterns than current pond connectivity, suggestive of possible effects of connectivity loss.


| INTRODUC TI ON
With escalating habitat loss worldwide, the conservation of small natural features is gaining increased focus (Bátori et al., 2014;Deák, Rádai, et al., 2020;Hunter et al., 2017). Temporary ponds and pools are naturally small, island-like habitats, with outstanding value for biodiversity conservation and ecosystem service provision (Céréghino et al., 2008;Davies et al., 2008;Hill, Greaves, et al., 2021;Lukács et al., 2013). They are viewed as model systems in conservation biology, ecology and evolutionary biology (De Meester et al., 2005). Due to the small size and scattered occurrence within the habitat matrix unsuitable for wetland species, these ecological islands are highly vulnerable and may require conservation instruments different from those applicable in larger wetlands (Blackwell & Pilgrim, 2011;Oertli et al., 2009;Semlitsch & Bodie, 1998). To develop effective conservation policies, ecologists are pressed to gain a better understanding of the mechanisms structuring their biological diversity Hill, Greaves, et al., 2021;.
The species-area relationship (SAR) has been used to evaluate the conservation value of island-like wetlands regionally Oertli et al., 2002) and to compare their biodiversity with other habitat types (Drakare et al., 2006;Matthews et al., 2016Matthews et al., , 2020. In temporary ponds, the strength of this relationship measured as the z value of the power (log-log) model varied largely across different systems, suggesting additional important factors involved in structuring pond diversity . A consistent external contributor to the diversity of ecological islands is the surrounding habitat matrix (Csergő et al., 2014;Martín-Queller et al., 2017;Matias et al., 2013). By varying the properties of the landscape matrix, e.g. through land use, we may facilitate or slow down the spread of pond habitat specialists, or of the undesirable weed and invasive species (Hartel et al., 2014;Nicolet et al., 2004;Oborny et al., 2007;Serrano et al., 2020). In particular, temporary ponds situated in or near floodplains might be heavily impacted by the dynamics of the surrounding habitat. Natural flooding regimes have maintained high productivity levels of floodplains by periodic water, nutrient and propagule supply, making them species-rich habitats and providers of abundant ecosystem services (Duncan, 2011;Schindler et al., 2016). As a result of the high appeal for agricultural use, 90% of the floodplains in Europe have been degraded and lost their functional role, while the remaining ones are experiencing distress syndrome manifest as lowered biodiversity and productivity, increasing prevalence of diseases and opportunistic and alien species (Rapport et al., 1985, Tockner et al., 2022. This chronic, sublethal state might have as well affected the species composition of the nearby temporary ponds (Lugo, 1978). As hydrological connectivity is key driver of both local and regional pond diversity (Akasaka & Takamura, 2012;Cohen et al., 2016), the restricted flooding regimes could cause pond hydroperiod changes with severe consequences on the persistence of different pond organisms (Rolls et al., 2018;Woolway & Maberly, 2020). Such fundamental system-level changes complicate the conservation of pondscapes, and yet conservation practitioners often lack a basic understanding of the spatio-temporal context of these habitats.
Even if species richness patterns were well known across a range of pond systems, emerging general conservation and restoration guidelines would be difficult if functional links between species and their environment were not recognized. Significant effort is being currently made in functional pond ecology to understand the interaction between organismal biology and community structure (Capers et al., 2010;Heino et al., 2015;Incagnone et al., 2015), and the determinants of functional trait variation in wetland environments (Iversen et al., 2022). In temporary ponds, a relatively high functional redundancy, that is, functional relative to taxonomic richness, was observed in aquatic invertebrates, along with a high beta diversity (Schriever & Lytle, 2016). This reflects the highly fluctuating hydric regime of the ponds and spatial constraints like low connectivity and small area, which may limit the diversity of species within the successful functional types (Heino et al., 2015;Scheffer et al., 2006). Therefore, analysing the diversity of functional groups may be more informative than the overall taxonomic diversity on local and spatial processes structuring pond diversity, and can more readily guide conservation decisions. Despite important advances in the field, the ecological and spatial drivers of pond functional group diversity, in particular of pond macrophytes, have remained insufficiently explored, delaying progress in pond conservation , but see Alahuhta & Heino, 2013.
We sought an easy functional group approach transferable across different systems to infer mechanisms underlying pond macrophyte diversity. Social behaviour types (Borhidi, 1995) describe planthabitat interactions within a framework extended from Grime's C-S-R evolutionary life strategies (Grime, 1988) and have direct utility in the conservation research of small natural features . In a model pondscape along the Olt river in the Eastern Carpathians, we grouped macrophyte species in different social behaviour types, and we investigated the effect of pond area and connectivity with the river on the diversity and abundance of these life strategies. We explored the potential effects of historical connectivity change on pond diversity by testing whether pond connectivity prior to river regulation and floodplain drainage explains better current pond diversity compared to the post-regularization environment.
We expected that (i) the diversity of different social behaviour types will exhibit specific responses to pond area and connectivity, which will differ in strength from the overall species' diversity; (ii) species richness, based on species presence, will be more strongly influenced by pond area and connectivity because it reflects better mechanisms involved in species colonization, while Shannon diversity and plant cover, estimated with species presence and abundance, will be less responsive to pond area and connectivity because are likely more influenced by local environmental conditions; and (iii) it is possible to link current pond macrophyte diversity to historical pond connectivity, and the river regulation and restricted flooding regime had differentiated effects on selected social behaviour types.

| Study site
The studied pond system is located in the Ciuc Basin, Eastern Carpathians, Romania (46.2461-46.4476 N;25.7302-25.8598 E, EPSG 4327 -WGS 84, Figure 1), a large tectonic basin characterized by alluvial-proluvial accumulation surfaces (Coteț, 1973) and cold and relatively dry climate (mean annual temperature: 3.8-7.6°C, total annual precipitation: 406-852.7 mm, Miercurea-Ciuc Meteorological Station, 1955-2002. The Olt River plays an important role in the hydrology of the basin, supporting a large floodplain environment as well as continuous wetlands, the extent of which increases from North to South along an approximately 100 m elevational loss (Rațiu, 1980). The river was drastically regulated in the 1970s and 1980s as part of a national programme on protection against floods and for gaining agricultural land, and a considerable network of wetland drainage structures was developed (ANAR, 2008). The landscape is open, with a mosaic of arable land and grassland used for extensive hay production.
Temporary ponds are a particular geomorphological and ecological feature of the area present in large numbers (currently about 260 are known), most likely of periglacial origin, formed by landslides on retreating permafrost at the Würm-Holocence transition (Figure 1, Kristó, 2002, Demeter et al., 2006. Some could rather be classified as wetland islands (Richardson et al., 2022), with an average surface area of 1116 m 2 , maximum depth usually smaller than 150 cm, over 30% vegetation cover and a highly fluctuating hydroperiod that includes regular dry phases in the second half of the year (Demeter et al., 2005 and unpublished data). However, due to the presence of temporary ponds and transitional types, in this paper, we maintain the term temporary ponds until further clarifications. The vegetation of the ponds does not show large seasonal or annual variation and likely develops at rather decadal or multidecadal timescales. Water pH and conductivity vary between ponds and between macrophyte communities within ponds (pH 4.7-6.7 and 23-382 μS, respectively), indicating high pond individuality and within-pond microhabitat heterogeneity (unpublished data). Algal biomass is low, and macrophytes form mainly marshland and fen and transitional mire communities, protected in Europe, which include glacial relict and rare plant and animal species (Bănărescu, 1995;Csergő & Demeter, 2011). The ponds are under anthropic pressure, some being mown in the dry phase, used as dumpsites or partially drained.

| Field data collection
We collected data on the macrophyte species diversity of 28 temporary ponds in the Ciuc Basin during peak vegetation season between 2006 June 9-11 (20 ponds) and 2009 July 13-18 (8 ponds), using a stratified sampling (Csergő & Demeter, 2011). The study ponds cov- one, 25 m 2 vegetation plot placed randomly within that community. Vascular species were identified at the species level and mosses at the genus level and assigned to six cover classes by visual estimation (Braun-Blanquet, 1964), then subsequently converted into percentages as follows: 0.5%, 5%, 17.5%, 37.5%, 62.5% and 87.5% (Cristea et al., 2004). In case of taxa difficult to identify in the field or identification uncertainties, specimens were coded, collected and subsequently,

| Species and functional group diversity and abundance
We grouped the recorded species into one of the social behaviour types (SBT) proposed by Borhidi (1995), except for two species present in one pond each with 0.5% cover, which were not found in this source. The social behaviour types along with example species are presented in Table 1. We used the relative "moisture figures" (WB) on the 12-grade F-scale of Ellenberg, also available in Borhidi (1995), to identify whether plants in the Specialists category were wetland plants according to the requirements for soil moisture.
To quantify plant diversity and abundance at the pond level and for each social behaviour type within a pond, we first calculated a pond-level pooled species dataset, selecting the average cover value observed for each species across all plots within a pond. Using the pooled dataset, we expressed species diversity as observed species richness (total number of species) and Shannon diversity index, and abundance as cumulative vegetation cover, calculated for each pond and each social behaviour type separately. To make the abundance values of different social behaviour types comparable, we divided the cumulative cover of each social behaviour type within a pond by the total vegetation cover obtained for the respective pond.
The diversity metrics were calculated using the vegan package in R (Oksanen et al., 2013). The affiliation of species to social behaviour types is available along with the vegetation plots in the Dryad data repository.

| Pond area and connectivity with the river
We recorded the geographic coordinates of each pond in the EPSG 4327 -WGS 84 Coordinate Reference System, using a GPS device.
We calculated pond area in 2006, using the short and long axes of the ellipse that best approximated the pond shape. We defined pond connectivity with the Olt River as the pond's present and, respectively, past geographic distance from the Olt River, measured between the pond's GPS coordinate and the nearest point of the river ( Figure 1). For obtaining the past distance, we used the digitized map of the Second Military Survey of the Habsburg Empire (1853-1858 and 1869-1870) accessed from https://www.arcan um.com (Timár et al., 2010). All geo-spatial operations were performed using QGIS Desktop 3.16.6. with GRASS 7.8.5 software. Pond-level data are available in the Dryad data repository.

| Statistical analyses.
We fitted linear models (LM) to test the effect of pond area and pond distance from the current riverbed on total species diversity, and diversity and abundance of each social behaviour type. When the number of species within a behaviour type was low across all ponds and missing from a large number of ponds, we instead modelled the probability of occurrence of those behaviour types with generalized linear models (GLM), specifying binomial error distribution. All models were subsequently refitted using pond distance from the river course prior to regulation. To evaluate whether the data collection year could have influenced the results, we refitted the same models with the data collected in 2006, which represented approximately 70% of the ponds.
Linear model residuals were checked for normality using the Shapiro-Wilk test and residuals plots, for heteroscedasticity with the studentized Breusch-Pagan test (lmtest package, Zeileis & Hothorn, 2002), and all models were tested for spatial autocorrelation of residuals using Moran's I index (ape package, Paradis & Schliep, 2018). When spatial autocorrelation was detected, linear models were updated with a spatial correlation term that produced the lowest AIC value using gls models (nlme package, Pinheiro et al., 2019), and for the binomial models, we fitted spatial GLM models (glmmfields package, Anderson & Ward, 2018). Species richness and pond area were log-transformed in all models, the relative cover of disturbance tolerants was log-transformed and the relative cover of specialists and generalists was logit-transformed except for one model of specialists in which log-transformation produced better model fit. To make the effect size of variables comparable across models, all continuous predictor variables were centred on 0 and scaled to have unit variance. All analyses were performed in R 3.6.0 (R Core Team, 2021). The structure of the models is presented in Supporting Material, Table S1 and Table S2. In the 28 sampled ponds, we recorded 164 plant species, with individual ponds having between 8 and 60 species, while pond average species richness was 27.3 ± 13.8 species and average Shannon diversity was 1.6 ± 0.3 (Table 1). The most diverse social behaviour types were the competitors, generalists, disturbance tolerants and specialists, the total number of which ranged between 32 and 43 species depending on the category. Across these categories, average pond species richness ranged between 5.6 and 8.1 species, Shannon diversity index between 0.75 and 1.17 and abundance values between 6.2 and 59.4% (Table 1). Natural pioneers, weeds and ruderal competitors were rarer, having less than 1.4 species on average, the mean Shannon diversity index lower than 0.3 and mean relative cover lower than 0.05%. We found no naturalized crops, adventitious weeds or alien competitors in the sampled ponds. Of the 34 Specialists, 27 were wetland ecosystem species that tolerate short or frequent floods, scoring 8 and above on the 12-grade F-scale of relative moisture figures by Ellenberg (results not shown).

| RE SULTS
Pond area had a significant, positive effect on overall species richness and the species richness of competitors, specialists and generalists, and it increased significantly the probability of occurrence of natural pioneers (log-log and binomial models, β > .237, R 2 > .158, p < .035; Figure 2a, Figure 3, Table S1). The z value of the species-area relationship was the highest for specialists (z = 0.413 ± 0.086), followed by all species (z = 0.296 ± 0.085), generalists (z = 0.272 ± 0.111) and competitors (z = 0.237 ± 0.073). Table showing the social behaviour types (SBT) in the plant strategy model by Borhidi (1995) and the number of species, mean (±1SD) species richness, Shannon diversity and relative cover (%) of each social behaviour type and all species across the 28 sampled ponds in the study area.  Note: Parentheses in the species richness column indicate minimum and maximum numbers of species found in a pond. Details of species richness, Shannon diversity and relative cover calculations are presented in the text. Note that two species could not be assigned an SBT. NA = not applicable.

TA B L E 1
Pond area had a significant, positive effect on the Shannon diversity of all species and competitors and specialists (semi-log models, β > .208, R 2 > .320, p < .003), a significant, negative effect on the relative cover of competitors (semi-log model, β = −.138, R 2 = .206, p = .015) and a significant, positive effect on the relative cover of Specialists (log-log model, β = 1.203, R 2 = .329, p = .001) (Figure 2a, Figure 3). There was no significant spatial autocorrelation between ponds in the species richness of social behaviour types except for the disturbance tolerants and occurrence probability of weeds (Table S1).
Pond current distance from the river had a significant negative effect on overall species richness, on species richness of specialists and generalists (log-log models, β < −.186, R 2 > .198, p < .042) and on Shannon diversity of specialists (semi-log model, β = −.194, R 2 = .320, p = .014) (Figure 2b, Table S1)  (Table S2). The z value of the species-area relationship was still higher for specialists (z = 0.382 ± 0.101), followed by competitors (z = 0.204 ± 0.090), all species (z = 0.155 ± 0.091) and generalists (z = 0.123 ± 0.122). The models retrieved all significant effects of pond's present distance from the river and detected additional significant, negative effects of this variable on the Shannon diversity of all species and generalists (semi-log models, β < −.128, Table S2). The models retrieved most significant effects of pond past distance from the river, except the effects on species' richness of specialists and disturbance tolerants (log-log models, β > −.086, R 2 = −.082, p > .501), and detected an additional significant negative effect of pond's past distance on the Shannon diversity of specialists (semi-log models, β = −.191, R 2 = .168, p = .039) (Table S2). We focus our discussion on the results based on the full dataset, with an outlook on the differences between the two sets of models. All model results are presented in Table S1 (for the full dataset) and Table S2 (for data collected in 2006).

| DISCUSS ION
Both pond area and connectivity with the river can explain the plant species diversity of temporary ponds in, and adjacent to, floodplains.
The varying strength of diversity and abundance responses of different social behaviour types indicate unequal ability of plants with different life strategies to colonize, persist and abound in temporary ponds of different sizes and connectivity, which can aid the identification of conservation targets. In the studied ponds, the connectivity loss with the river was relatively small, but we identified possible signals of connectivity loss on plant life strategy diversity within ponds.

| Diversity of plant life strategies in temporary ponds
Competitors, generalists, disturbance tolerants and specialists were the most common social behaviour types in the ponds. The high richness and cover of competitors indicate the productive nature of pond ecosystems (Grime, 1988) and the high richness and Shannon diversity of generalists and disturbance tolerants may be related to the pronounced hydric seasonality of these habitats (Biggs et al., 1994;Göthe et al., 2016), as well as abundant propagule supply from the landscape matrix. The relatively high richness, Shannon diversity and abundance of specialists reinforce that ponds are key contributors to the preservation of wetland species in agricultural landscapes and should be included in habitat conservation programmes (Csergő & Demeter, 2011;Lukács et al., 2013). The poor representation of weeds and ruderal competitors and lack of alien species indicated our preferential census of ponds without visible anthropic disturbance, and on the other hand, the favourable state of conservation of a large number of ponds in the studied area. Shannon diversity of all plant species, and species richness, Shannon diversity and relative cover of representative social behaviour types (competitors, specialists, generalists and disturbance tolerants) in 28 temporary ponds of Eastern Carpathians. Light grey points represent historical, and black points represent current pond distance from the river. Significant relationships detected by the linear models (LM) are indicated by fitted lines; confidence intervals are coloured in light grey if models were fitted with historical, and dark grey if models were fitted with current distance of ponds from the river. In (a), species richness plots are presented on log-log scale, and Shannon diversity and plant cover plots are presented on semi-log scale. In (b), species richness values are log-transformed. Y-axis values are comparable across the species richness, diversity and relative cover models respectively. In models of specialist species richness, we added one species to each pond for modelling purposes.

F I G U R E 3
The spatial distribution of species richness, Shannon diversity of all plant species and species richness, Shannon diversity and relative cover of representative social behaviour types influenced significantly by pond area according to the linear models (competitors, specialists and generalists) and generalized linear models (natural pioneers). Plots are not shown if the relationships were not significant in models corrected for spatial autocorrelation. Colour scales indicate the number of species, Shannon index and relative cover, respectively, and are not comparable between different plots. Dot size increases proportionally with pond area. Note the X-axis break for latitude.
in all types of ecological islands . While spatial processes may play a relatively important role in structuring plant communities of fluctuating wetland systems Capers et al., 2010;Incagnone et al., 2015), in our dataset the positive effect of pond area on plant diversity was likely not exclusively due to equilibrial dynamics of species. This is because, as postulated in the island biogeography theory, island area is often confounded with habitat heterogeneity (MacArthur & Wilson, 1967). Indeed, in our dataset, the number of species was positively correlated with the number of plant communities in the ponds (results not shown), suggesting increasing microhabitat heterogeneity with pond area.
As expected, we found weaker effect of pond area on the Shannon diversity of all species compared to the species richness. However, the effect was significant, positive and robust against differences in sample size, suggesting a strong control of pond area over abundance distributions of individuals in addition to the species number.
Therefore, the Shannon diversity index-area relationship can be a powerful indicator of the state of conservation of temporary ponds next to the conventional SAR. This pattern was paralleled only by the Shannon diversity of specialists and competitors, but while the relative cover of specialists increased, the relative cover of competitors decreased with increasing pond area. This result reinforces the antagonistic nature of the two life strategies (Grime, 1988), likely to respond differently to pond conditions underlying pond area. Of these, the hydrological regime is the most important in freshwater ecosystems (Rolls et al., 2018), and could explain the differences between the abundance of specialists and competitors.
The lack of a significant effect of pond area on disturbance tolerants, ruderal competitors and weeds indicates the ability of species in these groups to establish successfully in temporary ponds regardless of their size. This could become problematic for the conservation of small ponds if the propagule pressure of species in these less desirable categories increased as a result of incrementing land use intensity, which is an expected general tendency in this and other pondscapes (Csergő & Demeter, 2011;Oertli et al., 2009).

| Effect of pond connectivity with the river on plant life strategies
As expected, ponds closer to the river had higher total species richness and higher species richness of specialists and generalists. In our dataset, this beneficial effect may be due to higher soil moisture nearer to the riverbed having favourable influence on pond hydrology and enhancing the retention of wetland specialists, as well as to seasonal floods known to increase the propagule pressure in floodplains (Duncan, 2011). This result underscores the critical importance of hydrological connectivity in maintaining pond diversity (Akasaka & Takamura, 2012;Cohen et al., 2016). The steeper decline of specialists versus generalists Shannon diversity away from the river indicate constraints on the ability of species with narrower niche breadth to establish larger populations in remote ponds under limited connectivity. Specialists are more influenced by habitat limitations compared to Generalists, their success being more reliant on the dispersal ability, while generalists, due to larger niche breadths, are better able to overcome habitat and dispersal limitations (Büchi & Vuilleumier, 2016).
We recognize that pond distance from the river was an imperfect metric of connectivity because, as with many studies, we lacked knowledge of subsurface or surface-water connectivity and landscape features that may have enhanced the functional links between the studied ponds . However, even if we underestimated pond connectivity, the effect of the simple geo- The average pond distance from the river had decreased only slightly after the regulation of the Olt River, and as a result, we observed only minor differences in the effects of pond distance from river on species' diversity. Because we lacked pond biodiversity data prior to river regulation, when the existence of these ponds was barely known (Demeter et al., 2005), we cannot ascertain the true effects of the modification of the river course. However, if past connectivity is indeed reflected in current biodiversity patterns, and our results seem to confirm this hypothesis, then we detected a recent increase in the control of the river on the Shannon diversity of specialists and a lowered control on the species richness of disturbance tolerants, Shannon diversity of generalists and relative cover of competitors. These would be among the first signs expected under a steeper soil moisture gradient away from the floodplain, and lowered intensity of floods due to faster water drainage. Increasing environmental distress observed in floodplains worldwide puts a strain on particular types of organisms, while others are less affected (Woolway & Maberly, 2020). Our results show that modified diversity patterns of a few social behaviour types could represent early warning signals of system-wide changes in the hydric regime of pondscapes. The diversity patterns in our pondscape have likely developed over multidecadal or even centennial timescales due to complex hydrological dynamics and macrophyte colonization and persistence histories (Cohen et al., 2016;Demeter et al., 2005;Kristó, 2002). The full effects of human interference with the topology of the ponds will thus be slow to develop, with far-reaching consequences not only on the biotic composition of the ponds but also on the water and biogeochemical cycles and ultimately, on the continued provision of ecosystem services in the studied landscape (Cohen et al., 2016).

| CON CLUS IONS
Our results demonstrate the utility of functional island biogeography approaches Deák, Rádai, et al., (Borhidi, 1995;Grime, 1988) can provide a more refined understanding of how pond area and connectivity shape pond plant composition compared to general measures of total taxonomic richness. Declining richness, diversity and relative abundance of wetland specialist species can represent early warning signals of system-level changes in the state of conservation of ponds and focusing efforts on their monitoring could be more effective than thorough taxonomic evaluations. Further, while pond area and associated habitat heterogeneity exert a strong control on plant diversity, and therefore, are an important predictor of biodiversity, the effects of landscape-level factors such as connectivity cannot be ignored when managing temporary ponds for conservation. The maintenance of natural processes in the adjacent landscape matrix should therefore be an integral part of pond management Murphy & Lovett-Doust, 2004). In exchange, temporary ponds can be sensitive gauges of the landscape condition.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data used in this manuscript will be made freely available on Dryad upon the publication of the manuscript (URL: https://doi. org/10.5061/dryad.2jm63 xss9).

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13685.