The influence of climate warming on flowering phenology in relation to historical annual and seasonal temperatures and plant functional traits

Introduction

Since the late 19th century the average global temperature has risen by 1.1 °C (IPCC, 2021), and in the contiguous United States, the yearly average temperature has increased by 0.09 °C each decade since 1895, surpassing the 1895–2020 average in 1956 (NOAA, 2021). In the region in which this study was conducted, the yearly average temperature increased by 0.11 °C each decade since 1895, surpassing the 1895–2021 average in 1958 (NOAA, 2022). This change in climate has the potential to influence many biological and ecological processes, particularly changes in phenology (timing of biological events) in plants and animals, given the potential for temperature-dependence of such traits (Walther et al., 2002). Moreover, the effects of climate change on phenology may in turn have further implications in terms of the consequential impact on ecological interactions. Among these is the potential impact of climate change on pollination due to the differential temporal response of plants and their pollinators to warming seasonal temperatures resulting in asynchrony between flowering time and the emergence of pollinators (Liu et al., 2011; Rafferty & Ives, 2012; Robbirt et al., 2014; Forrest, 2015; Rafferty, 2017; Renner & Zohner, 2018). Similarly, a differential response to warming between prey emergence and predator demand has been shown to significantly impact different trophic levels in the same community (Harrington, Woiwod & Sparks, 1999). Additionally, independent shifts in the phenology of invasive plants and their herbivores in response to warming suggest that climate change will influence biological invasions (Walther et al., 2002; Walther et al., 2009; Hellmann et al., 2008).

Given the potential for climate warming to influence plant flowering phenology and the broader ecological consequences of such changes, this has become an important area of study. The effect of warming on flowering phenology has been addressed experimentally through the physical warming of plots in natural settings (Price & Waser, 1998; Memmott et al., 2007; Flynn & Wolkovich, 2018; Zettlemoyer, Schultheis & Lau, 2019; Dorji et al., 2020; Pérez-Ramos et al., 2020; Collins et al., 2021); in field studies over temperature gradients with elevation (Hoiss, Krauss & Steffan-Dewenter, 2015; Inouye, 2020; Richman et al., 2020) and latitude (Parmesan, 2007; Bertin, 2008; Niesenbaum, 2012; De Manincor et al., 2020), and along urban-rural temperature gradients resulting from heat island effects in cities (Neil & Wu, 2006). Another approach has been to use historical data from herbarium collections to assess the effects of warming on flowering phenology. Herbarium specimens are most often collected during blooming, indicating flowering dates and location, and many such collections span hundreds of years, offering a unique long-term historical perspective of the relationship between flowering time and historical trends in warming (Meineke, Davis & Davies, 2018).

The potential limitations of using herbarium specimens to document long-term changes in flowering time stem from collection biases due to non-random collecting practices related to location, sampling time, and collector preferences for certain species, as well as duplicate sampling of individual species from the same location on the same date (Robbirt et al., 2011; Daru et al., 2018; Willis et al., 2017; Gamble & Mazer, 2022; Yang et al., 2022). Although these biases can limit the value of phenological information of individual specimens, the use of large sample sizes while avoiding duplicates; and including plants from a large number of families, and with different functional traits and flowering seasons can allow for a more accurate assessment of the long-term shifts in phenology at a given location (Willis et al., 2017; Daru et al., 2018). Bias can also come from inconsistent assessment of phenophase or the timing of onset, peak, and end flowering dates, which becomes more problematic for species that flower over longer parts of a season. This can be mitigated by using a consistent application of flowering time based on the percentage of open flowers, and by using plants with similar durations of flowering (Davis et al., 2015). Finally, the use of the collection date of flowering herbarium specimens as a proxy for actual flowering time has been validated in studies that show no significant difference between specimen collection dates and flowering dates observed in the field in the same year (Bolmgren & Lönnberg, 2005; Miller-Rushing et al., 2006; Robbirt et al., 2011; Jones & Daehler, 2018; Ramirez-Parada, Park & Mazer , 2022; Bloom, O’Leary & Riginos, 2022).

Prior studies on long-term trends in flowering phenology in relation to historical temperatures revealed changes in flowering time by as much as 2–6 days earlier per 1 °C increase in temperature (Primack et al., 2004; Panchen et al., 2012; Park & Schwartz, 2015; Davis et al., 2015), and 8–16 days earlier per 1 °C in more recent decades (Primack et al., 2004; Lavoie & Lachance, 2006). However, not all species show the same temporal response, with some blooming earlier and others later over time (McEwan et al., 2011; Calinger, Queenborough & Curtis, 2013; Hart et al., 2014; Davis et al., 2015; Bertin et al., 2017; Panchen & Gorelick, 2017). One possible reason for observed interspecific differences in response to warming is that individual species may respond differently to seasonal temperatures within a year, such that winter or spring temperatures could independently or divergently influence flowering time depending on species specific cues for germination, and leaf or flower emergence (Cook, Wolkovich & Parmesan, 2012; Pearson, 2019). Many prior studies of climate warming on flowering phenology have been based solely on the temporal trends using collection year as the only explanatory variable. While this may be well suited for examining long term trends in phenology with warming, since temperatures have risen during the collection periods; linking collection dates to concurrent temperature data from the same locations allows for the consideration of seasonal temperatures within a year as drivers of phenological change. Use of temperature data can also more accurately reveal year-to-year variation in phenological response to temperature (Holle, Wei & Nickerson, 2010; Panchen et al., 2012; Bertin, 2015; Davis et al., 2015; Park & Schwartz, 2015; Daru et al., 2018; Ramirez-Parada, Park & Mazer , 2022; Zettlemoyer, Wilson & De Marche, 2022).

Another reason for mixed results among species in prior studies of warming on flowering phenology may be due to differences in plant functional characteristics known to be related to phenology (Parmesan & Hanley, 2015). Interspecific character differences that likely influence temporal responses to warming include native status (native versus non-native species), growth habit (woody versus herbaceous species), fruit type (species with dry versus fleshy fruit), and seasonality of blooming (spring-blooming versus summer-blooming species). The native status of species could influence phenological responses to temporal warming because non-native species, especially those that are invasive, tend to exhibit more phenotypic plasticity than native species (Richards et al., 2006; Davidson, Jennions & Nicotra, 2011), and thus may be more likely to exhibit changes in flowering time in response to warming temperatures especially over the short term (Calinger, Queenborough & Curtis, 2013; Wolkovich et al., 2013; Zettlemoyer, Schultheis & Lau, 2019). Growth habit also has the potential to influence the temporal changes in flowering time. There is some evidence that in woody species phenology may tend to be cued by photoperiod more than temperature, so woody species may be less likely to exhibit changes in flowering time with temperature deviations (Saikkonen et al., 2012; Zohner et al., 2016; Flynn & Wolkovich, 2018). Fruit type may influence the phenological response to warming since plants with fleshy fruits may be more constrained by fruit and seed development time and disperser abundance than timing for pollination (Primack, 1987; Du et al., 2020), which in turn could influence the response to warming. Also, accounting for the season in which different species bloom can shed further light on interspecific divergent responses to warming (Pearson, 2019).

The objective of this study was to use herbarium specimens for 36 species from 28 families collected from 1884–2015 to determine general patterns of long-term changes in flowering time in relation to concurrent historical temperature data in eastern Pennsylvania, USA. We also assessed whether yearly average temperature, and winter and spring temperatures elicit different responses across species. Additionally, our goal was to determine the extent to which differences in native status, growth habit, fruit type, and the seasonality of blooming are related to the response of flowering phenology to warming. We predicted that flowering time will tend to occur earlier in response to warming over the time period considered, and that there would be differences among species in the predictive strength of yearly average, and winter and spring temperatures on flowering phenology trends. We also hypothesized that the relationship between flowering phenology and warming temperatures would be stronger in non-native species in contrast to native species, in herbaceous compared to woody species, in species with dry fruits relative to those with fleshy fruits, and in spring blooming species compared to those that flower in summer.

Methods

Results

The linear mixed-effects regressions for all species with individual species as the random effect were significant for annual and spring temperatures, but not winter temperatures, as indicated by whether or not the 95% confidence intervals included the value of 0. These analyses revealed that from 1884–2015, plants in this study on average bloomed approximately 2.26 days earlier per 1 °C increase in the yearly average temperature (β = −2.26, 95% CI [−3.27 to −1.26], Fig. 1). Winter temperatures did not significantly influence flowering phenology (β = −0.16, 95% CI [−0.59 to 0.38], Fig. 2), while plants flowered approximately 2.93 days earlier per 1 °C increase in the spring onset average temperatures (β = −2.93, 95% CI [−3.62 to −2.27], Fig. 3). When we excluded species with extended blooming periods beyond five months the effect was slightly stronger for yearly average (β = −2.72, 95% CI [−3.96 to −1.74]), and spring onset temperatures (β = −3.16, 95% CI [−3.83 to −2.38]) and remained non-significant for winter temperatures (β = 0.09, 95% CI [−0.39 to 0.55]). There was no significant difference in the effect of temperature on flowering phenology between native and non-native species for any of the three temperature periods (Table 2). The comparison between woody and herbaceous growth plant species showed a significant difference in flowering phenology only in response to the yearly average temperature, with woody species (slope = −4.53 d/° C) flowering 2.85 days earlier than herbaceous species (slope = −1.68 d/° C) with increasing temperatures (95% CI [−4.919 to −0.822]) (Table 2, Fig. 4).

For the comparison of fruit types, there was no significant difference in flowering phenology in response to warming annual, winter, or spring temperatures (Table 2). The comparison between spring and summer blooming species was significant only for yearly average temperatures (Table 2, Fig. 5). Spring blooming species exhibited a significantly greater response to warming (slope = −3.284 d/° C) than summer blooming species (slope = −1.014 d/° C), with the difference in slopes revealing that spring blooming species flowered 2.27 days earlier per 1 °C increase in yearly average temperature than summer blooming species (95% CI [0.485 to 4.662]). The exclusion of species with extended flower periods did not influence any of the results of the categorical analyses in Table 2.

Discussion

We have shown that in eastern Pennsylvania flowering phenology has significantly advanced in response to rising temperatures. The species in this study bloomed on average approximately 2.26 days earlier per 1 °C increase in the yearly average temperature from 1884–2015 (Fig. 1). This is consistent with other similar studies showing plants flowering 2–6 days earlier (Primack et al., 2004; Miller-Rushing et al., 2006; Robbirt et al., 2011; Anderson et al., 2012; Panchen et al., 2012; Calinger, Queenborough & Curtis, 2013; Hart et al., 2014; Bertin, 2015; Davis et al., 2015; Park & Schwartz, 2015; Bertin et al., 2017; Pearson, 2019; Williams et al., 2021; Bloom, O’Leary & Riginos, 2022; Yang et al., 2022). The plants in this study are on the low end of this range for earlier flowering with warming. One possible explanation for this is that much of the area of study remained as undeveloped agricultural land through the early 1990s (Lehigh Valley Planning Commission, 2022) and none of the specimens used in this study were collected from large urban or industrial centers, perhaps in contrast to other areas in which this type of study have been conducted. This may have reduced any heat island effect thereby amplifying warming trends and is known to influence flower phenology (Neil & Wu, 2006; Zipper et al., 2016). Another potential issue is the general decline in herbarium collection rates in more recent decades which could result in an under estimate of phenological response to warming in more recent years. This is the case for a few of the species in our study, which we attempted to address through the use of temperature rather than year as an independent variable in analyses. However, it is possible that this resulted in an underestimation of the of the phenological response to warming in more recent years. Also consistent with prior studies showing early flowering with warming for a number of species, is that not all species flowered earlier and some actually flowered later (Cook, Wolkovich & Parmesan, 2012). Furthermore, differences in the plant species between studies of this sort could explain variation in results due to possible phylogenetic constraints on the phenological responses to warming.

To further explore this variation among species and to shed additional light on the possible mechanistic bases of changes in flowering time with climate change, we also considered the influence of seasonal temperatures within a year and a number of plant functional characteristics that could potentially influence the response of flowering time to warming. In our analysis of the influence of seasonal temperatures on flowering time across all species considered here, we found that warmer spring temperatures were a strong driver of phenological response with plants on average flowering 2.93 days earlier per 1 °C (Fig. 3), which was greater than the response to increases in annual temperature (Fig. 1). This makes sense given the conventional thinking that earlier spring warming would initiate the termination of dormancy, germination, and leaf and flower emergence; and is consistent with prior studies that examined this (Primack et al., 2004; Miller-Rushing et al., 2006; Robbirt et al., 2011; Calinger, Queenborough & Curtis, 2013; Davis et al., 2015; Bertin et al., 2017; Park et al., 2018). However, the question remains why some plant species appeared not to respond to warming temperatures.

The lack of response of some species to warming average may actually be driven by warmer winter temperatures which can delay dormancy or chilling requirements for germination via vernalization, which in turn could delay spring events such as flowering or leafing out in some species (Hart et al., 2014; Rubin & Friedman, 2018; Cook, Wolkovich & Parmesan, 2012). This is consistent with our finding that increasing winter temperatures were not significantly related to flowering phenology, and the lack of a negative slope in this relationship, suggests that some species may actually be flowering later in response to warming winter temperatures (Fig. 2). This supports the suggestion by Cook, Wolkovich & Parmesan (2012) that species that appear not to respond to warming average annual temperatures seen in this and other studies may actually be responding to warming, but the opposing effects of winter and spring warming cancel each other out. Sampling error and limited sample sizes may also be a possible explanation for the varied response among species.

We expected the native status of species to influence phenological responses to temporal warming because of the potentially greater phenotypic plasticity of non-native species (Richards et al., 2006; Davidson, Jennions & Nicotra, 2011), but we found no difference between native and non-native species in the relationship between flowering time and rising annual and seasonal temperatures (Table 2). Other studies that considered this found similar or mixed results (Holle, Wei & Nickerson, 2010; Bertin, 2015; Bertin et al., 2017), with only one other that found non-native species to flower earlier in response to rising temperatures (Calinger, Queenborough & Curtis, 2013).

There are a number of possible explanations why we did not see non-native species flowering earlier in response to warming than natives. First, not all introduced species are invasive, and whether or not a plant is invasive may be a more important determinant of plasticity than native status. This however is difficult to discern as most studies of plasticity in non-native plants focus on those that are invasive. Since only species with specimens that span the time range over which warming has occurred and that the rate of introduction has increased exponentially during the time of this study (Lehan et al., 2013), the proportion of non-native species that are invasive may have been under represented in this and similar studies. In fact, only five of the non-native species used in this study, Alliaria petiolata, Datura stramonium, Hesperis matronalis, Lonicera japonica and Lythrum salicaria, and Lonicera japonica (ital) (Table 1) are designated as invasive (USDA NRCS, 2021a). If invasive non-natives exhibit greater phenotypic plasticity than those that are not invasive, the underrepresentation of invasive species in our study could explain the lack of response to warming of the non-native species seen here and in other studies. Also, most studies of invasive plants assess plasticity characters such as reproductive output, growth rate, and photosynthesis (Richards et al., 2006), and plants with greater plasticity may also respond to warming by investing more in early vegetative growth. Finally, adaptive selection on phenological advance also plays a significant role in response to warming (Anderson et al., 2012), potentially equalizing differences in the response of native and non-native plants attributed to plasticity. If we are to better understand how phenological changes may influence biological invasion with climate warming, we will need to advance our understanding of the underlying mechanisms.

We found that the relationship between flowering time and warming was significantly stronger in woody plants than that in herbaceous plants, contradicting our own hypothesis. We are unaware of other studies that directly make this same comparison. There is some evidence that in woody species phenology tends to be cued by photoperiod rather than temperature; but this tendency varies within particular growth forms, and with latitude, flowering season, and other factors (Saikkonen et al., 2012; Zohner et al., 2016; Flynn & Wolkovich, 2018). Thus, to better predict the consequences of climate warming on flowering phenology, direct comparisons among species with different growth forms and life histories are needed, as is a more comprehensive understanding of the relative effects of temperature and day length on phenology across species. Our finding of no differences in phenological response between species with dry or fleshy fruits (Table 2) is also in conflict with our own hypotheses and prior work (Bolmgren & Lönnberg, 2005). This may be due to the low number of fleshy fruited species in our study, or the possibility that fruit type may or may not be related to flowering time or other functional characteristics.

We found that spring-blooming species show a greater advance in flowering phenology than summer-blooming species. Most past studies found the same (Calinger, Queenborough & Curtis, 2013; Bertin, 2015; Bertin et al., 2017; Pearson, 2019), with Park et al. (2018) finding the opposite. The greater advance in flowering time with warming of spring blooming species can simply be explained by the earlier cue of warming on phenological processes. Also, selection on spring blooming plants to flower earlier to decrease competition for pollinators, particularly for male function, is typically balanced by the risk of late winter damage (Niesenbaum, 1992; Austen & Weis, 2016). As that risk decreases with climate warming, selection to flower earlier should become stronger (Calinger, Queenborough & Curtis, 2013). Pearson (2019) demonstrated this effect in a much warmer climate where spring blooming species exhibited a stronger phenological response to warmer spring temperatures than species that bloom later in the season; and that significantly warmer summer temperatures actually delayed flowering in summer and autumn flowering species. Additionally, Pearson (2019) found that the phenological advance in spring flowering was dampened in years with high levels of spring precipitation independent of temperature.

Conclusion

It is clear from our study and others like it that climate change is influencing flowering time in ways that could negatively impact community level ecological processes. However, it also clear that the phenological responses to warming varies greatly among species due to the multiple factors and mechanisms at play. These include differential responses to winter and spring warming, and different functional characteristics among those species. Future work should focus on these possible mechanisms through both historical and experimental studies that take into account the complexities of seasonal differences in warming, and multiple factors that likely contribute to the variation in the response of individual species and how they might interact or be confounded with each other. Consideration of other factors related climate change such as seasonal precipitation, snow melt, the frequency and intensity extreme weather events, and comparisons among sites with greatly different climates will further enhance our understanding of how climate change influences flowering phenology (Lavoie & Lachance, 2006; Anderson et al., 2012; Peterson, 2016; Flynn & Wolkovich, 2018; Hufft et al., 2018; Dorji et al., 2020). Additionally, analysis of species specific factors that cue flowering phenology (Davis et al., 2015), the consideration of the influence of urban heat island effects related to urban and suburban development (Roetzer et al., 2000; Ziska et al., 2003; Lavoie & Lachance, 2006), potential interactions among the plant characteristics that we considered, as well as taking into account any phylogenetic signal in flowering response to warming could help to make sense of observed differences in the phenological response to warming.

Supplemental Information