Urbanization and plant diversity influence different aspects of floral phenology

Plant phenologies can shift as global temperatures rise and landscapes become human-dominated. Some of the most pronounced shifts in phenologies have been documented in urban areas, where surface temperatures can reach 5–7℃ warmer than surrounding rural areas. In this study, we examined how floral phenology—the initiation, peak, and duration of floral events — shifted in grasslands across an urbanization gradient in Louisville, Kentucky, a city with one of the most severe urban heat islands in the US. Our objectives were to understand (1) how urbanization influences floral phenology (2) whether high-quality habitats with increased habitat patch size and plant richness could offset some effects of urbanization, and (3) whether species responses varied across seasons. We found that average first date of flowering and peak abundance date occurred 1–2 weeks earlier in urban compared to rural areas. However, we found that floral duration was longest in sites with high plant richness, regardless of urbanization. We also found that summer-flowering species increased their floral duration in urban areas while spring and fall-flowering species shortened theirs. These differences in seasonal responses lead to an “urban summer spillage” effect where summer-flowering species are able to move into the temporal niche of spring and fall species in urban habitat patches. These shifts could lead to a reshuffling of communities and novel plant-plant competitive interactions or plant-animal interactions in the urban core, with lasting implications for urban conservation.


Introduction
Urban areas are expanding and intensifying globally as more of the human population moves into cities (Seto et al. 2012). This urban expansion and fragmentation of the natural landscape can lead to dramatic abiotic changes, local extinctions, reduction in plant and animal genetic diversity, introduction of alien species, and disruptions to species interactions (Radeloff et al. 2005, Johnson andMunshi-South 2017). Among the most pronounced and consistent communities. It is important, therefore that we continue to understand better the effects of urbanization on floral phenology of forb communities and the context dependencies of this relationship. Factors such as site quality, landscape context, and species identity likely play a significant role in how urbanization influences plant phenologies.
Understanding species-specific or guild-specific responses to the urban heat island is particularly important because variations in responses could lead to novel interactions and reshuffling of communities (CaraDonna et al. 2014). By changing the abiotic environment so drastically, urbanization will constrain, or expand the functional niche of many species, but also has the potential to influence the realized niche by generating novel species interactions. If a shift occurs in the phenology of a species particularly susceptible to temperature changes, urban heat-islands may open temporal niche space for another to dominate. Identifying the groups of species that will strongly respond to urban heat-islands will be important for understanding how urban communities are reshuffled. For example, urban heat-island responses may be particularly strong for springflowering species. Several studies in both urban and natural settings have documented stronger responses to elevated temperatures in spring-flowering species compared to those that flower in summer or fall (Neil and Wu 2006;Ding et al. 2020), including in a review of 385 British species (Fitter and Fitter 2002). There are several possible explanations for why spring species may respond so strongly to temperature changes. Spring-flowering species may be at heightened risk for phenological mismatches with pollinators, and so may be more plastic in their responses to temperatures to avoid reduced seed set (Kudo et al. 2004;Kudo and Ida 2013). Spring species also experience heightened costs from a mistimed emergence from frost damage, compared to summer and fall species, which drives physio-and morphological selection (Rodrigo 2000). These costs and pressures may lead to increased selective pressure on spring species beyond morphology and into their phenologies, making them more responsive to changes in the abiotic environment than summer or fall species.
Fortunately, local site characteristics related to habitat quality have the potential to offset some of the deleterious effects of urbanization on plant phenology. Urban parks and other islands of natural vegetation can reduce local temperatures by reducing impervious surface cover (Norton et al. 2015). Studies have also suggested that larger green spaces may be able to buffer against the heat island in more interior sections as vegetative complexity increases (Rafiee et al. 2016;Cheung and Jim 2019). Additionally, the plant neighborhood, or community composition, may influence floral phenology. Grassland plants in higher diversity systems have increased drought resistance and decreased pest damage (Tillman and Downing 1994). In these diverse systems plants may be able to reallocate resources away from defense and towards reproduction, and therefore shift their floral duration, peak date or initiation.
Our study set out to understand how urbanization and habitat quality interact to influence floral phenology. Specifically, we ask (1) Are landscape-level characteristics of urbanization associated with shifts in plant community floral phenology? (2) Can local site quality factors offset some of these shifts in phenology? And finally, (3) Are species' responses to urbanization consistent across seasons? We predicted that we would see plants flowering earlier and for shorter periods in urban habitats and smaller habitats compared to the rural habitats and larger habitats. We also predicted that increases in local site quality could reduce effects of urbanization. Finally, we hypothesized that while many species may shift their phenologies in response to urbanization, the change would be greatest for spring-flowering species.

Study system
In 2018 and 2019, we conducted floral surveys in small grassland restorations in and around the city of Louisville, Kentucky, United States (38.253° N, 85.759° W). Annual precipitation in Louisville was 170 cm in 2018 and 116 cm in 2019 (NWS Louisville weather station, NOAA), with summertime temperatures between 25 and 35°C, and a growing season typically lasting from April through October. Louisville's metro population is roughly 1.3 million, and the urban core is one of the strongest heat islands in the USA, with average warm season high temperatures and cool season low temperatures 5-7° C warmer than the surrounding rural areas (Stone et al. 2019). We surveyed 17 sites in 2018 and 13 in 2019, all of which had been actively managed as native grasslands ( Fig. 1; Geographic coordinates and land cover values available in our external repository). Restorations ranged in age from 1 to 15 years old, with an average size of 4.22 ha (0.5-15 ha range). We removed five sites from sampling in 2019 because of changes in site management, and in 2019 we added one new restoration site. There was no correlation between site size and urbanization (measurement details below; r 2 = 0.06, p = 0.31), nor urbanization and plant richness (r 2 = 0.10, p = 0.25), as other studies focused on urbanization and pollinators have found (McIntyre 2000), which allowed us to look at the combined effects of these factors on floral phenology.

Phenological data collection
From April -October of 2018 and 2019 we surveyed the floral community every other week at each site, for a total of 13 surveys per site per year, 390 surveys across the entire study. For surveys, we randomly selected a location in the middle of each grassland site and marked this location with a cavity-nesting bee nest (used for a concurrent study, see Sexton et al. 2021). From this point, we ran four 20 m transects radiating out in each cardinal direction. Along each transect, four 2 × 2 m sampling subplots were spaced every 5 m, and all inflorescences were identified to species and counted during each survey. Floral species richness values from each survey were averaged to give one whole-season metric of floral richness per site per year. We calculated three metrics of floral phenology for each species in each site: relative floral initiation date, relative floral duration, and relative peak flowering date. These measures allow for a more nuanced understanding of phenology compared to just using floral initiation date (Austen et al. 2014; Inouye

Statistical analyses
To address how univariate floral phenology metrics responded to local and landscape-level factors we used multi-model inference, or AIC model averaging, as outlined in Burnham and Anderson (2004). In this approach, key predictor variables were identified by comparing all possible combinations of predictors using the coefficients from each model. The summed model weights range from 0 to 1, with higher values indicating that the predictor is more likely to be a part of the well-supported models (by AIC score).
We also separated each of our univariate phenology responses (relative floral initiation, duration, peak) by season of flowering (spring, summer, fall) and sampling year (2018, 2019), and ran each model with three predictor variables of each site: urbanization, size, and floral species richness. Data from different years were analyzed separately due to differences in sample sizes. All statistical analyses were performed in R software, version 4.1.0 (R Development Core Team 2019).

Results
We recorded  While we saw strong effects of urbanization, floral richness also strongly influenced floral phenology. In both 2018 and 2019 floral duration was extended by roughly three weeks in sites with higher floral richness than in low et al. 2019). At the end of each field season, we also categorized plant species by whether they were spring-(first appearance in surveys 1-4), summer-(surveys 5-9) or fall-(surveys 10-13) flowering to calculate season-specific phenology metrics.
We calculated relative floral initiation date by subtracting the flower initiation date of each species at a specific site by the average initiation date across all sites for a given species. This measure provides insight on whether species tend to flower earlier or later at specific sites compared to the region as a whole. The same was done for duration, where we subtracted the duration (# of surveys present) of a given species at a given site by the average duration for that species, and for peak flowering date. For these metrics we excluded all species that only occurred at one or two sites, as our research question was focused on how species change from site to site across a gradient. From this restriction, we used 63 species for phenological analysis of the 124 species recorded in the field in 2018 and 48 of the 64 species in 2019.

Landscape measures
We quantified two landscape factors that may influence floral phenology in our system, habitat size and urbanization, following methodology used in Sexton et al. (2021). We used ArcMap 10.6 to measure habitat size by drawing polygons around site edges using aerial imagery (LOJIC Metro 2016 3-inch Map). To measure urbanization, a 1.5 km radius buffer zone around the center points of each of our sites was selected, and surrounding land cover values for each buffer zone were determined using data from the 2016 National Land Cover Database (Jin et al. 2013). A buffer size of 1.5 km was selected as this is the buffer size used in similar studies and is the typical flight range of a large bee, making it relevant for a study on floral phenology (Zurbuchen et al. 2010, Bennett andLovell 2019). Land classes from this dataset included developed land (Open, Low, Medium, and High combined into a single category here) and 11 classes of vegetative cover (Deciduous Forest, Coniferous Forest, Agriculture, etc.). We conducted a principal component analysis (PCA) to create a single composite measure of urbanization based on the 12 classes of land usage (loading scores in Appendix S1). PC1 explained 74% of the variation in land usage, and the variable loading scores on the first principal component (PC1) indicated an urbanization gradient. PC1 has high positive values for Developed Land, and strongly negative values for Forested Land. For further analyses, PC1 scores were used as the continuous measure of landscape-scale urbanization gradient. PC1 was also strongly correlated with % impervious surface from a separate GIS layer (USA NLCD Impervious), further but consistent trends are beginning to emerge, at least at the landscape scale. Several remote sensing studies have documented earlier 'green-up' as well as expanded summer growth in urban areas compared to surrounding rural areas, indicating a change in plant phenophases (Dallimer et al. 2016;Yuan et al. 2020;Li et al. 2017b). Our study complements this landscape-level work by demonstrating that floral phenologies of individual species within plant communities are similarly responding to urbanization.

Seasonal variation in phenological responses
Summer-flowering species exhibited the positive response to urbanization, extending their flowering duration in urban areas while spring and fall species shortened theirs. Additionally, in 2019 we saw spring species shifting their flowering peak earlier in urban areas. These phenological shifts indicate an "Urban Summer Spillage" effect, whereby summer flowering species are able to expand into the temporal niche of spring and fall-flowering species in urban areas (Fig. 3). This indicates that in urban areas spring and fall ephemerals face a reduction in their functional niche due to rising temperatures, but also a potential reduction in their realized niche due to novel competitive interactions with summer-flowering forbs. These novel competitive interactions cover several aspects, including competition for pollinators, nutrients, and moisture. Additionally, specialization in plant-pollinator networks tend to be highest in the spring and when resources are scarcer (CaraDonna and Waser 2020; Souza et al. 2018). As more generalists are introduced into what were previously spring communities, specialist interactions may face heightened strain in urban areas. Finally, urban summer spillage may also contribute to homogenization of floral communities in urban areas. This richness sites (2018 -ME: 0.161, SE: 0.033 p < 0.001; 2019 -ME: 0.194, SE: 0.074, p = 0.009). When broken down by seasonality, this effect was consistent for spring species in both 2018 (ME: 0.195, SE: 0.049, p = 0.001) and 2019 (ME: 0.284, SE: 0.121, p = 0.02). In 2018, summer species also extended their duration in the high richness sites (ME: 0.308286, SE: 0.088, p < 0.001), while fall species were not affected either year (all full model outputs stored in Appendix S2). In 2018 we also saw floral initiation advance by approximately one week in high richness sites across the entire dataset (ME: -0.077, SE: 0.031, p = 0.01). Site size did not have any measurable effect on any metric of floral phenology. There was a marginal (p = 0.08) effect of larger sites extending the duration of spring flowering species, but again no significant effects.

Urbanization impact on phenology
We saw strong effects of urbanization on multiple parameters of phenology, including floral initiation, peak, and duration. Across all species, urbanization was correlated to earlier initiation and peak flowering date, likely as a result of the urban heat island effect. The temperature gradient along Louisville's particularly strong urban heat-island is greater than the temperature shifts examined in many other studies that simulate climate-related warming by manipulating temperatures by just 2-3°C (Stone et al. 2019). Pronounced shifts in floral phenology as a result of climate warming have been well documented in recent years, especially in montane/alpine systems (Inouye 2008). Research on how urbanization influences floral phenology is less extensive, Bars indicate +/-1SE. Positive coefficients indicate an increased value of each response (longer duration, later peak dates, later initiation date) as each predictor value increases and may be able to invest more heavily in reproduction, thereby increasing floral duration.
It is also possible that site diversity is not directly affecting phenology but is instead an indicator of disturbance regimes or habitat quality. While we were unable to obtain detailed management records of our sites, it is possible that greater management attention in the form of burning or mowing may alter phenological patterns. Mola and Williams (2018) found that floral phenology was lengthened in grasslands that had been recently burned, leading to increases in floral abundances in summer months.
Changes in soil microbial communities associated with increased plant diversity may also be responsible for phenological shifts. Hahl et al. (2020) found that soil microbial communities originating from plant monocultures caused plants to shift their investment to defense-related traits, and away from growth and reproduction. Other greenhouse and field studies have shown that rhizospheric organisms can have significant influences on floral phenology (Wagner et al. 2014;Lu et al. 2018). Arbuscular mycorrhizal communities can increase total floral production, and floral rewards, i.e., nectar and flower size (Barber and Gorden 2015). Future studies investigating how feedbacks between plant community diversity and soil microbial communities, can influence floral phenology are warranted.
Our study found no evidence that habitat patch size influences floral phenology or modulates effects of urbanization. We expected that larger sites would be more buffered from heat island effects, as several studies have shown that even minimal vegetation increases, as much as 10-50% within 20 m radii, can reduce surface temperatures by 1.0°C in urban parks (Coutts and Harris 2013;Cheung and Jim 2019). It may be that our study was not able to detect size effects because we worked in relatively large patches, in regards to urbanization. It is likely that site size may play a larger role in reducing the effects of urbanization in smaller habitat patches such as community gardens and residential lawns. Finally, due to limitations in human-power and a large geographic area, our sampling interval was relatively long compared to some other phenology studies that are able to sample daily. This means it is possible we may have missed species with particularly short flowering periods. Investigating responses of these short-duration species in urban areas is another area for future interest.

Conclusion
Floral phenology is a multi-faceted phenomenon, and here we show that different aspects of floral phenology, including floral initiation, peak, duration, and schedule shape, can shift independent of one another in response to environmental phenomenon has been observed in other urban areas where distant urban plant and animal communities are more like each other than they are to the surrounding rural communities (McKinney 2006;Groffman et al. 2014).

Local site factors influence phenology
While our work shows that urbanization has significant impacts on floral phenology, we also found evidence that local site characteristics can directly influence phenology as well. Plants flowering in richer plant communities exhibited extended periods of floral duration compared to populations of the same species in low-richness communities. Several other studies have documented increased plant performance in high-diversity grassland communities. For example, more diverse grassland communities have been found to be more drought resilient, produce increased biomass, and have more resistance to invasion than low-diversity communities (Tilman and Downing 1994;Kreyling et al. 2017;Hahl et al. 2020). It is possible that in more-diverse grassland communities, plants may experience reduced stressors