Weather and climate may influence the phenology of Dirca occidentalis A.Gray (Thymelaeaceae) in ways that impact reproductive success. Dirca occidentalis blooms during winter, when the likelihood of entomophily may be low. Based on preliminary observations that the timing of dormancy release and growth resumption varies over years and among shrubs within years, we quantified fruit set among flowers that formed at different times and examined whether annual variation in autumnal precipitation and temperature during autumn and winter are associated with phenology. Fruit set was determined during 2007–2008 through 2011–2012 by tracking 37,461 flowers near or at anthesis early, midway, and late within the blooming period of D. occidentalis. In addition, measures of phenology of 18 individual shrubs were made each December 29, January 26, and February 23 of the five blooming periods, and fruit set of these shrubs was determined. Fruit set was low (<5%) among flowers present early (December 26–January 2), but increased significantly in all blooming periods, to as high as nearly 30%, among flowers at anthesis later. Phenology ratings, and lengths of newly formed stems and leaves, on December 29 increased linearly as the amount of precipitation from October 1–December 15 of the same year increased. Phenology ratings on February 23 increased linearly with increasing air temperature from November 1–February 23. Rankings of phenology of the 18 shrubs were highly correlated over years, and fruit set of individual shrubs over years was 1% to 52% and increased linearly as growth resumption and flowering became later. Our results demonstrate that low autumnal precipitation is associated with delayed growth resumption and flowering, which corresponds with increased fruit set of this rare species.
Plant phenology is closely associated with weather and subject to changes in climate (Parmesan and Yohe 2003; Menzel et al. 2006a, b; Cleland et al. 2007; Willis et al. 2017; Yost et al. 2020). As sessile organisms, plants are robust indicators of their environments. Phenophases of plants are strongly affected by interannual variability in weather and climatic patterns, and changes in environmental factors influence vigor, competitiveness, and survival (Walther et al. 2002; Davis et al. 2015). Because temperature and water relations are among the most important determinants of plant phenology (Menzel et al. 2006a, b; Gill et al. 2015), year-to-year thermal and hydric variation can alter the timing of phenological events (Menzel et al. 2006a, b; Cleland et al. 2007; Inouye 2008).
Regions with Mediterranean climates, such as coastal California, are characterized by warm, dry summers and cool, wet winters. Some plants in Mediterranean climates undergo drought deciduousness in summer. Leaves desiccate and may abscise as plants become quiescent during extended dry periods before rainfall in autumn or winter promotes the resumption of shoot growth (Griffin 1973; McCreary 1990). Annual variations in weather can shift the onset of growth resumption, and the timeframes of growth (McCreary 1990; Prieto et al. 2008) and flowering (Wolkovich et al. 2013). Earlier flowering in winter and spring has been associated with elevated cool-season temperatures and autumnal precipitation (Prieto et al. 2008). While phenological asynchrony between pollinators and their food plants caused by climate change is not the norm (Hegland et al. 2009; Forrest 2014), temporal mismatches between some pollinators and early-flowering plants have been documented (Memmott et al. 2007; Tylianakis et al. 2008; Renner and Zohner 2018). Such misalignments can reduce pollination and fecundity. Species distributed narrowly in small populations may be particularly prone to extirpation amid climate change (Aitken et al. 2008).
Dirca occidentalis A.Gray, a member of the only extant genus of the plant family Thymelaeaceae indigenous to the continental United States, is endemic to the California counties of San Mateo, Santa Clara, Contra Costa, Alameda, Marin, and Sonoma (Graves and Schrader 2008). Known as leatherwoods, Dirca spp. are shrubs with arborescent forms, precocious yellow flowers in late autumn through early spring, flexible stems, and fibrous bark with high tensile strength (Hudson 2019).
Among the species of Dirca, only D. occidentalis occurs in a Mediterranean climate. Plants typically are clustered in the understory of sloped woodlands, but chaparral (Ackerly 2004) and semi-riparian habitats also may include the species. Reproduction is primarily sexual (Graves and Schrader 2008). Copious hermaphroditic flowers are borne annually, but the single-seeded drupes of D. occidentalis are rarely observed (Johnson 1994). Because shrubs bloom during winter, low temperatures may limit the activity of insect pollinators, a phenomenon observed in other members of the genus (Williams 2004; Graves 2008). The phenology of D. occidentalis has received little attention, however, and our preliminary observations indicate considerable variation among individuals. As a narrowly endemic species in a family that is not prevalent in the flora of North America, D. occidentalis merits research to improve our understanding of its phenology and how climate change may affect the species.
Dirca occidentalis is found in numerous habitats within Jasper Ridge Biological Preserve (JRBP) in San Mateo County. The species is most abundant in evergreen woodlands, but also is found in riparian areas along creeks, scrub, and chaparral. At 61.6 to 211.5 m above sea level, the 483-hectare preserve covers the northern half of Jasper Ridge within the San Francisquito Creek watershed. Data from local weather stations show average annual precipitation from 1975 to 2004 was 652 mm. About 90% of annual precipitation occurs from November through April; the warmest months are August and September, when mean temperature is about 17.5°C; and the coolest months are December and January, when mean temperature is about 10.5°C.
Because of summer deciduousness, the functional lifespan of leaves of D. occidentalis in chaparral communities at JRBP is four months (Ackerly 2004). Dirca occidentalis we observed in woodlands, scrub, and chaparral at JRBP became quiescent during summer. Further, the timing of phenological events marking growth resumption appeared asynchronous among individuals, and fruit set appeared to vary markedly among plants and over years. Based on these preliminary observations, our objectives were to examine whether annual variation in autumnal precipitation and temperature was associated with phenology, and to quantify fruit set among flowers that formed at different times within the long blooming period of the species. Our approach was to quantify fruit set by tracking tens of thousands of flowers that formed early, midway, and late in the annual blooming period for the species over five consecutive winters. In addition, the phenological statuses of the same 18 plants were recorded on the same three dates during those five winters, and the fruit set of those plants was determined. We provide evidence that the timing of growth resumption among individual shrubs following summer quiescence is highly variable, that precipitation and temperature are strongly associated with phenology of D. occidentalis, and that differences in phenology are closely associated with variation in fruit set.
Fruit set of D. occidentalis was documented during five consecutive winters, 2007–2008 through 2011–2012. Numerous locations throughout JRBP were scouted each winter to include D. occidentalis in the range of habitats in which the species occurs. Most D. occidentalis were found in the understory of wooded, north- and east-facing slopes, but individuals also were adjacent to waterways at the bottoms of slopes, and were scattered within upland chaparral communities. Dataloggers (iButton model DS1921G, Maxim Integrated, San Jose, CA) were used to record air temperature at 120-minute intervals 1.5 m above grade at 10 locations chosen to represent the habitats of the D. occidentalis included in this study. Although D. occidentalis in all habitats were included each winter, the individual plants on which fruit set was documented varied among winters and were not tracked. Flowers on shrubs of various sizes were tagged. Because shrubs of average size usually support copious flowers, only small fractions of the total flowers per shrub typically were tagged.
Within each winter, inflorescences were tagged to allow subsequent determination of fruit set. Only inflorescences with flowers that had developed at least to the stage of pistil extension (Stage 3, Fig. 1) were tagged, and no inflorescences with flowers past anthesis were tagged. Most tagged inflorescences comprised flowers at anthesis (Stage 4, Fig. 1), with extended pistils and stamens, and anthers that had begun to shed pollen. Tagging was done within one-week periods early (December 26–January 2), midway (January 23–30), and late (February 20–27) during the flowering period for the species to determine whether fruit set differed during the three periods when flowers were tagged. Air-temperature data during each of these weekly periods and the week immediately following them were obtained from the dataloggers to encompass the timeframes of anthesis. Averaged over the five blooming periods, mean temperatures during these three two-week periods were 6.6°C, 8.1°C, and 8.7 °C, respectively, with maximums of 14.0°C, 16.8°C, and 18.3°C. Among the five blooming periods, only during 2010–2011 were temperatures higher in the middle week (mean = 9.3°C, maximum = 22.6°C) than during the late week (mean = 8.0°C, maximum = 19.1°C).
Tags were pieces of adhesive-tape 2–3 cm long. A tag was attached to the stem immediately below the node of each chosen inflorescence. Care was taken to avoid contact with flowers during tagging, and to position tags such that they could not touch flowers. Tags were removed when data were recorded after approximately one month. Because flowers that do not set fruits abscise one to two weeks after anthesis, fruits were readily noted as swelling ovaries. Data on fruit set, determined as the percentage of flowers that led to a developing fruit, were analyzed to determine effects of the five blooming periods and the individual weeks of tagging flowers.
Data were fit by using a generalized linear model (GLM) in RStudio (R version 3.5.2; Rstudio, PBC, Boston, MA). Estimated dispersion with the binomial family of the GLM function was 0.726, so the quasibinomial family was used (dispersion parameter of 0.891). Estimated marginal means were calculated with the emmeans package, and predicted probabilities along with their 95% confidence intervals were plotted (Fig. 2). A compact letter display of all pairwise comparisons was calculated using the cld function of the multcomp package. Results were plotted with symbols that indicate air temperature during flowering by using the ggplot2 package.
Inflorescences of Dirca occidentalis Tagged (and the Corresponding Number of Flowers) Monitored, to Document Fruit Set. Tagging was done during three one-week periods in five consecutive flowering periods. The three weeks were during early (December 26–January 2), middle (January 23–30), and late (February 20–27) portions of the flowering period of the species.
Thirty plants of D. occidentalis that represented the diverse habitats in which the species occurs at JRBP were chosen to track phenology during the winters of 2007–2008 through 2011–2012. Twelve of these plants were disregarded during the study because they died or were damaged by felled trees or herbivory significantly enough that obtaining data over five winters was not possible.
The 18 remaining plants were rated each winter to document their developmental statuses on December 29, January 26, and February 23. A seven-level scale was used, where 1 denoted a quiescent plant, 4 denoted full anthesis, and 7 denoted that flowering had ended, fruits were developing, and that newly formed stems had elongated substantially (Fig. 1). In addition, the lengths of newly expanding leaf blades and elongating stems were measured to the nearest 0.1 mm by using a Verner caliper. After thoroughly examining the developing tissues, what appeared to be the five longest leaves and stems of each plant were measured.
Fruit set of these plants was also estimated. Rather than tagging inflorescences, approximately one month after the floral period of a plant ended, 30 floral nodes along stems were checked for the presence of developing fruits. The 30 floral nodes were chosen in the same way for each plant. First, the tallest branch of the shrub was identified. Then, the uppermost floral node along that branch was checked, followed by the 29 floral nodes that immediately subtended it. Whether a node was floral was determined easily because each flower present within an inflorescence resulted either in a fruit or in a distinctive scar on the stem where the pedicel had abscised. The number of flowers that the 30 inflorescences had comprised was determined as the number of fruits plus the number of scars, and fruit set was calculated as the percentage of flowers that led to a fruit. The 18 plants were ranked annually based on mean ratings (Fig. 1) on the three dates from 1 (earliest to develop, leading to the highest mean phenology rating) to 18 (latest to develop, leading to the lowest mean phenology rating). Means of the five annual mean rankings for each plant were used in a linear regression analysis to examine whether mean ranking was associated with mean fruit set over the five blooming seasons.
Environmental data from 2007 to 2012 were obtained from the 10 dataloggers and a weather station at JRBP. Total precipitation from October 1 to December 15 was used in regression analyses to determine whether precipitation was associated with the mean phenological rating, and mean leaf and stem lengths, of the 18 plants on December 29. The relationship of air temperature and phenology was examined with a linear regression analysis of mean temperature from November 1 to February 23 and mean phenology rating of the 18 plants on February 23.
Means of the three phenology ratings recorded each year for the 18 plants was determined to examine their relationship to fruit set. Because residual plots showed heteroskedastic variance of error, fruit-set data were square root-transformed, which corrected the non-constant variance. The function lmer() from the R package lme4 was used to fit a mixed linear model to this set of repeated measures collected over the five annual blooming periods. Random intercepts for plants were assumed independent, and Restricted Maximum Likelihood was the estimation method. In addition, Spearman rank correlations determined by using the function cor() from base R were used to assess consistency in how the 18 plants ranked from year to year based on their mean phenology ratings.
Ample flowers were available to tag during each of the three weeks of all five blooming periods. Total inflorescences tagged per weekly period ranged from 500 to 1096, which corresponded to 1519 to 3419 flowers (Table 1). The total number of inflorescences tagged per annual blooming period ranged from 2,159 to 2,666. The overall total number of flowers included in this study across the three weekly periods and five annual blooming periods was 37,461.
During most blooming periods, fruit set was higher among flowers blooming during the middle week (January 23–30) and late week (February 20–27) than in the early week (December 26–January 2) (Fig. 2). The only exception to this was in 2007–2008, when there was no difference between the early and middle week. Across all five blooming periods, mean fruit set was <5% for flowers present during the relatively cool early week, whereas mean fruit set was highest among flowers blooming in the late week during three of the five periods (Fig. 2). In 2008–2009, however, there was no difference in mean fruit set between the middle and late weeks, and, in 2010–2011, fruit set was highest in the middle week. The highest mean fruit sets, nearly 30%, were in the middle week of 2010–2011 and in the late week 2011–2012 (Fig. 2).
Among the 18 plants that could be monitored during all five blooming periods, mean phenology rating on December 29 increased linearly with seasonal precipitation from October 1 through December 15 (Fig. 3). For each 1-cm increase in precipitation, mean phenology rating increased by 0.009 (R2 = 0.91, P < 0.05). Leaf expansion and stem elongation also increased linearly with total precipitation from October 1 through December 15 (Fig. 3). Mean length of the longest leaves increased by 0.105 mm per 1-cm increase in precipitation (R2 = 0.82, P < 0.05), and mean length of the longest stems increased by 0.070 mm per 1-cm increase of precipitation (R2 = 0.92, P < 0.01). Mean phenology rating on February 23 increased linearly (R2 = 0.88, P < 0.05) by 1.44 per 1°C increase in temperature (Fig. 4).
Differences in phenology of these 18 plants were associated with their fruit set. Mean fruit set over blooming periods increased linearly (R2 = 0.49, P < 0.01) with later flowering and vegetative development as indicated by increased mean phenology rank (Fig. 5). The mixed linear model showed an effect of mean phenology rating (P < 0.001) and the fixed effect of year (P < 0.001). The variance of the plant random effect was 0.6213 with a residual error of 1.5694. The estimate for the effect of phenology rating was –0.8817 (t = –6.482, df = 35.8, as determined by Satterthwaite approximation). The adjusted intraclass correlation coefficient (ICC) was 0.284.
Spearman rank correlations of yearly rankings of the 18 plants based on their mean phenology ratings were strong and positive among all 10 pairwise combinations of years. Correlations ranged from 0.77 to 0.94, and the median of the correlations was 0.90.
This research provides new information regarding the autecology of D. occidentalis. First, our data provide new insights regarding variation in the phenology of this species. Individual shrubs varied in phenology within blooming periods, and phenology of the population of shrubs at JRBP collectively shifted across years. The fact that flowering precedes the resumption of vegetative growth in late autumn or winter in D. occidentalis accounts for the second insight from this work: the variation in the timing of flowering is associated with differences in fruit set. Third, there is evidence that rainfall and air temperature are associated with phenology and fruit set.
The numerous shrubs of D. occidentalis within the 483-hectare JRBP bloom asynchronously over approximately four months. Individual shrubs did not bloom throughout that period. Rather, flowering was highly synchronous within individual shrubs that were floral during different portions of time from December through February. The phenological statuses during winter of the 18 shrubs we monitored over five blooming periods were highly variable. For example, their statuses on January 26 ranged from Stage 1 to Stage 7 (Fig. 1) in 2009, 2010, and 2012, and the ranges were Stage 1 to Stage 6 in 2008 and Stage 2 to Stage 6 in 2011. Individual shrubs were floral (Stages 3–5 or 6) for approximately two to three weeks, and individual flowers were at anthesis for up to one week. The type of asynchronous flowering we observed was described as plastic by Rodríguez-Pérez and Traveset (2016). Most shrubs of D. occidentalis provided flowers in particular abundance in late January and late February (Table 1), but we were able to tag at least 1519 flowers each year from December 26–January 2, and we observed flowers several weeks before late December in most years, as well as flowers after late February in all years.
Although clusters of plants within some habitats tended to bloom rather synchronously, we frequently observed one to a few outlier plants within larger clusters that did not flower at the same time as the majority. Based on such observations, and the fact that D. occidentalis can spread clonally (Graves 2004), future studies could examine genetic relatedness of individuals within clusters and whether the degree of floral synchrony among closely positioned individuals influences fruit set. As the image depicting Stage 3 of our phenology rating scale illustrates (Fig. 1), D. occidentalis may be protogynous. Results of pilot studies we have conducted on pollination suggest the species is self-compatible, but that cross pollination bolsters fruit set and seedling vigor. Further research is needed to document the pollination biology of D. occidentalis and the consequences of the mode of pollination on fruit seed, seed viability, and seedling vigor.
The timing of annual resumption of growth and flowering among individuals of D. occidentalis varies for unresolved reasons that may include genetic variation (Graves and Schrader 2008) and microsite variation in air temperature and edaphic characteristics that influence soil water content. Pratt and Mooney (2013) observed intraspecific variation in phenophase timing of Artemisia californica Less. provided different quantities of water in a garden plot. Such plastic asynchrony of flowering may provide a degree of resistance to effects of climate change on reproductive success. The strongly positive Spearman rank correlations provide evidence for consistency across years in the degree of earliness of individuals. We noted that the same individual shrub ranked as the earliest to pass through the phenological stages we defined (Fig. 1) each blooming period, and only two different shrubs were ranked second earliest across the five years. In addition, the same shrub ranked latest of the 18 in four of the five years. Greater variation in rankings was evident for individual shrubs that were neither extremely early nor late to develop. Evaluation of the range of factors that may influence annual variation in phenology among individuals of D. occidentalis merits additional research.
Based on our preliminary observations that D. occidentalis at JRBP is floral over long periods comprising late autumn and all of winter, and a report that fruits rarely form (Johnson 1994), we questioned whether time of anthesis is related to fruit set. We conclude that it is. Among large samples of flowers tracked by tagging them while at anthesis, fruit set was at least tenfold higher for flowers tagged during the late week than for flowers tagged during the early week, and the greatest percentage of fruit set was from flowers tagged during the middle or late week within each of the five annual blooming periods (Fig. 2). Consistent with these findings, we observed that fruit set increased linearly among the 18 individuals we monitored as phenology of annual growth resumption became later (Fig. 5). The negative effect of phenology rating when fruit-set data were analyzed with a mixed linear model further demonstrates increased fruit set with later flowering and resumption of vegetative growth. The adjusted ICC from the linear mixed effect model indicates fruit set of individuals is modestly positively correlated across years, i.e., a low to moderate tendency for individuals to be consistent in their fecundity.
As a species that is putatively entomophilous, the increase in fruit set with later flowering could be linked to increased pollinator activity later in the blooming period (Corbet et al. 1993). Air temperature, which often is strongly associated with the time of flowering (Abu-Asab et al. 2001), tended to be lowest during weeks when flowers led to few fruits (Fig. 2). Pollinator activity could explain this observation. Low and mean temperatures we recorded were below the threshold that Williams (2004) associated with pollinator visits to D. palustris, which flowers in late winter and early spring in eastern North America. In contrast, relatively high fruit set was coincident with high temperatures during the middle week of 2010–2011 and during the late week of all other blooming periods (Fig. 2). We therefore posit that individuals of D. occidentalis that resume annual growth and bloom relatively late are more likely to set fruits than shrubs that serve as harbingers by resuming growth and blooming relatively early, i.e., before mid to late January.
How precipitation can influence pollinator activity also merits consideration. Precipitation events when flowers are at anthesis are most likely during relatively wet blooming periods, which tended to hasten flowering of D. occidentalis (Fig. 3). By damaging flowers and disrupting pollinators, rainfall, along with low temperature, might suppress fruit set. Water from rainfall can alter the osmotic potential of pollen grains, causing them to burst (Corbet and Plumridge 1985; Lawson and Rands 2019), and pollen may not adhere to and germinate on wet stigmatic surfaces (Ortega et al. 2007). Rainfall also disrupts pollinator activity by removing floral volatiles, physically limiting pollinator movement, and altering nectar rewards through dilution (Tot-land 1994; Cnaani et al. 2006; Lawson and Rands 2019). Furthermore, the activity of small pollinators that putatively service Dirca is suppressed during periods of low solar radiation typical of precipitation events (Willmer 1983; Kilkenny and Galloway 2008). The interplay of environmental factors, particularly precipitation, air temperature, and solar radiation, on insects that pollinate D. occidentalis merits elucidation.
In addition to variation among individual plants in the phenology of annual growth resumption and flowering, this research also provides evidence that the temporal window of dormancy release, shoot development, and flowering of D. occidentalis shifts from year to year at JRBP. The extent of annual shoot development evident on December 29, first manifested by swelling buds (Fig. 1), was positively correlated with the amount of precipitation from October 1 to December 15. Regression analyses predict increasing phenology ratings, greater extension of newly formed stems, and longer newly developing leaves with increasing autumnal precipitation (Fig. 3). The influence of precipitation on floral phenology varies among species (Ogaya and Peñuelas 2004; Peñuelas et al. 2004; Crimmins et al. 2010) and may differ regionally (Abu-Asab et al. 2001). In a study of two Mediterranean shrub species that bloom in autumn, Prieto et al. (2008) found that an imposed drought treatment in late spring and summer delayed flowering of Globularia alypum L., while there was no effect on Erica multiflora L. Precipitation was more predictive than temperature of the timing of annual vegetative development of three of four species of woody plants in California analyzed by Mazer et al. (2015). Their results demonstrate the importance of precipitation timing. For example, the release of pollen from Baccharis pilularis DC. was hastened and delayed by precipitation during December and February, respectively, while, in Quercus lobata Née, bud break was either promoted or delayed by precipitation, depending on the month during winter when it occurred. In a study of A. californica, which is more broadly distributed than D. occidentalis, Pratt and Mooney (2013) found that changes in water availability advanced and delayed flowering depending on the provenance along a 700-km gradient from which plants had been cloned. Based on these studies, additional research should build on the relationship we found between precipitation from October 1 to December 15 and the phenology of D. occidentalis (Fig. 3). Additional research should further define how the amount and timing of precipitation affect dormancy release and flowering, and the roles of temperature and soil edaphic conditions that influence shrub water status. Based on our results, we hypothesize that the annual resumption of shoot growth of D. occidentalis is triggered relatively early during blooming periods with high precipitation in autumn (Fig. 3), and that the progression through stages of seasonal development (Fig. 1) accelerates as temperatures in November through February increase (Fig. 4). A robust test of this idea would comprise shrubs from genetically distinct populations on the peninsula south of San Francisco, the East Bay, and the North Bay (Graves and Schrader 2008).
This research provides the first data documenting the phenology of the rare, endemic shrub, D. occidentalis. Our findings demonstrate marked phenological variation among individual shrubs, differences in fruit set based on floral phenology, and the association between environmental factors and phenology. Although this research provides valuable new information, additional research on the fate of initiated fruits is needed. The fraction of these fruits that mature and contribute viable seeds that germinate and give rise to new individuals remains unknown, as is how climate change will affect D. occidentalis. If autumns and winters in the Bay Area of California become wetter and warmer, our results suggest that flowering of D. occidentalis would be hastened, potentially resulting in lower fruit set and recruitment of new individuals unless there are simultaneously shifts in the activity of effective pollinators. In contrast, while drier autumns and winters might delay the flowering period and increase fruit set, reduced water availability in winter and early spring could suppress seed germination and seedling establishment, as well as the vigor of mature shrubs.
We are grateful to the staff of the Jasper Ridge Biological Preserve (JRBP), particularly Nona Chiariello and Cindy Wilber, for support and assistance; David Ackerly for insights on D. occidentalis at JRBP; and Miranda Tilton for consultations on statistical analyses.