With abundant evidence of recent climate warming, most vegetation studies have concentrated on its direct impacts, such as modifications to seasonal plant and animal life cycle events (phenology). The most common examples are indications of earlier onset of spring plant growth and delayed onset of autumn senescence. However, less attention has been paid to the implications of continued warming for plant species' chilling requirements. Many woody plants that grow in temperate areas require a certain amount of winter chilling to break dormancy and prepare to respond to springtime warming. Thus, a comprehensive assessment of plant species' responses to warming must also include the potential impacts of insufficient chilling.
Evidence of recent climate warming, as well as its impacts on the biosphere are abundant (IPCC, 2007). Phenology, the study of seasonal plant and animal life cycles, driven by environmental change ‘…is perhaps the simplest process in which to track changes in the ecology of species in response to climate change’ (IPCC, 2007). Numerous phenological studies have identified important direct effects of warming, such as earlier onset of spring plant growth and delayed onset of autumn senescence across temperate mid-latitude climates (Parmesan and Yohe, 2003; Root et al., 2003; Menzel et al., 2006; Schwartz et al., 2006). Yet, many woody plant species growing in these regions also paradoxically require a certain amount of exposure to cold temperatures (termed the fall/winter chilling requirement), in order to properly break their dormancy and be ready to respond to springtime warming (Sarvas, 1974; Cannell and Smith, 1983; Chuine and Cour, 1999; Baldocchi and Wong, 2008; Luedeling et al., 2009).
Thus, with recent and projected future warming, a point will eventually be reached (likely different for every species), where plants in temperate climates will no longer be able to continue linearly expanding both ends of their growing season. It is highly likely that temperate plant species in many regions, especially those on the warmer (southern) extremes of their range, may already be inadequately chilled and no longer responding in the same way to additional spring warming (Zhang et al., 2007). Therefore, comprehensive assessment of likely plant species responses (growing season change) to continued warming in temperate mid-latitude climates must include the potential impacts of insufficient chilling (Baldocchi and Wong, 2008; Luedeling et al., 2009).
When necessary chilling is not received, more springtime warming is required to initiate budburst (Murray et al., 1989). The direct effect of this on a plant species should be slowing or even a suspension of the rate at which the onset of the growing season advances towards earlier dates, with additional climate warming. Perhaps more importantly, process-based phenology models suggest that lack of chilling also has long-term consequences for some species viability, through reduced reproductive success (Morin et al., 2007). Lack of chilling may also decrease plant productivity, thereby reducing carbon reserves needed to overcome stress from pests or extreme weather events. Thus, being able to assess chilling adequacy has important implications for the future ecology and vulnerability of many temperate plant species.
Phenological studies have provided primary evidence of the direct impacts of climate warming on the temperate biosphere. Now, we contend that, when recorded at a continental-scale, plant species' phenological data can also be used to extract information relating to the large-scale impacts of warming (and reduced chilling) on plant species' physiology. A recent study supports this contention, and asserts that some southern parts of the United States are already experiencing the impacts of the chilling inadequacy described above (Zhang et al., 2007). However, these and similar studies that rely on remotely sensed information cannot explore this issue as thoroughly as surface species-based phenological observations (Schwartz, 1998). In this study, we demonstrate how first leaf and first bloom phenology from multiple locations across the western United States (when matched with temperature records) can estimate a plant species' chilling requirement and evaluate the changing impact of warming on the plant's phenological response in light of that requirement.
2. Data and methods
We obtained common lilac (Syringa vulgaris) first leaf (660 cases/84 stations) and first bloom dates (1428 cases/111 stations) recorded over the 1967–2003 and 1956–2003 periods respectively at National Weather Service Cooperative (COOP) stations distributed across the western United States (from 32.10 to 48.78°N, 99.73 to 121.17°W, and elevations of 119 to 2404 m; Figure 1), and combined these with co-located daily maximum–minimum air temperature records (Cayan et al., 2001; Schwartz and Caprio, 2003; NCDC, 2005). The first leaf event is defined as occurring when the widest part of the newly emerging leaf has grown beyond the ends of its opening winter bud scales (inset within Figure 2). The first bloom event is defined as occurring when at least 50% of the flower clusters (a grouping of many small individual flowers) have at least one open flower (inset within Figure 3).
We evaluated the impact of chilling levels on first leaf and first bloom dates by first calculating accumulated chilling hours (below a 7.2 °C base temperature, commonly used for horticultural trees and shrubs; Linvill, 1990; Schwartz, 1997) and chill units received between 1 October and 1 February each year from the air temperature records. The chill unit method assigns variable weights (determined by a sinusoidal curve) to chilling hours associated with different temperature levels, such that temperatures of 0 °C or less are set to zero, temperatures above 0 °C increase in value to a maximum of + 1 at 7 °C, then begin decreasing, reaching zero again and becoming negative after 14 °C, eventually reaching a minimum of − 1 at temperatures higher than 25 °C (Richardson et al., 1974; Linvill, 1990; Table I). We then matched these chill hour and chill unit values with the corresponding phenology data at the same locations. For all of the resulting phenology/chilling data sets (first leaf and first bloom), we next applied K-means clustering (two-cluster solution) to determine the best breakpoint in the relationship between phenological date and chilling accumulation. Lastly, linear regression equations were calculated (and residuals assessed) for both resulting clusters (in all data sets) in order to quantify the differing relationships between chilling accumulation and phenological data. We used scatter plots to graphically display the overall relationships.
Table I. Chill unit calculation for each hour spent at a specific temperature (constant = π × 2/28, Sine is computed in radians; Linvill, 1990)
0 °C or less
Chill unit = 0
Greater than 0 °C, up to 25 °C
Chill unit = Sin (constant × temperature)
Greater than 25 °C
Chill unit = − 1
In order to provide additional information to put the results in context, we also calculated 1961–1990 average annual accumulated chilling hours (received between 1 October and 1 February) for COOP stations in the study area (with at least 25 years of data available over this 30-year period), and examined individual stations for any trends in chilling hour accumulation over the 1956–2003 period. Further, we calculated the average timing (calendar date) of the relationship breakpoint between phenological date and chilling hour accumulation, as well as its variation by year and location, over the 1956–2003 period.
Scatter plots, K-means clustering, and regression analyses did not indicate any strong or significantly changing relationships among chill unit accumulation and first leaf/first bloom date; therefore, no further analyses were conducted with chill units. The K-means clustering analyses did indicate a break in the relationship between chilling hour accumulation and phenological date at 1748 chilling hours for both the first leaf and first bloom phenology/chilling data sets (Figures 2 and 3). For station/years that reached this chilling accumulation level, the 1956–2003 study area-wide average date that it was achieved was 13 January, with averages at specific locations varying from 21 December to 20 April, and the overall study area yearly averages ranging from 28 December to 14 February.
The regression equations showed changes for both events in the relationship between accumulated chilling hours and phenology (dates get earlier as chilling gets less) for the clusters containing chilling accumulations above the 1748 threshold level (slopes from linear regression equation of − 5.0 days/− 100 chilling hours for first leaf and − 4.2 days/− 100 chilling hours for first bloom). For the clusters of both events with chilling accumulations below that level, the rate at which the phenology advances is greatly reduced (slopes from linear regression of − 1.6 days/− 100 chilling hours for first leaf and − 2.2 days/− 100 chilling hours for first bloom), as suggested by the chilling theory (previously discussed). All regression slope values were significant as per the t-test at α = .0005. Residual distributions for all regression equations were not significantly different from normal, with α values >.10 as per the K–S test with Lilliefors significance correction. None of the residuals showed any coherent spatial patterns; however, all four sets were correlated with their corresponding phenological dates.
The average study area-wide chilling accumulation over the 1956–2003 period was 2235 h, with overall yearly averages varying from 2082 to 2415 h. Chilling accumulation varied much more rapidly in the southern portions of the study region (Figure 4). A small number of locations (28) showed significant changes in chilling accumulation over the 1956–2003 period, with some indication of regional trends toward less chilling in Washington State, California, and Arizona, and more chilling in interior portions of the study area.
The residual correlations suggest that follow-up work should consider non-linear options for describing the relationship between chilling accumulation and first leaf/first bloom date. Our results agree with much more extensive studies that show on-going reductions in chilling accumulations (projected to continue throughout this century) within the fruit growing regions of California (Baldocchi and Wong, 2008; Luedeling et al., 2009). Yet, possible chilling accumulation trends in other regions may also be worth examining in detail, along with an assessment of the impact of topography on this variable.
Overall, the encouraging results for common lilac suggest that similar continental-scale phenological measurements could facilitate a better understanding of relationships among chilling requirements, phenological response, and springtime warming for other species. Further, such data, because they would provide plants species' responses across large portions of species geographic ranges, will facilitate deeper understanding of the full range of plant–environment responses and consequently foster the development of more robust phenological models.
As we have shown how important chilling considerations are for evaluating the impacts of likely future warming on plant species, several important follow-up plant ecology questions should also be addressed in future studies using the same general methodology: (1) Are the chilling requirements for a species the same across its entire range? (2) Do species adapt to warming conditions by changing their chilling requirements? (3) How much variation is there among species chilling requirements within the same community? Such investigations, as facilitated by continental-scale phenological data sets being developed by the USA National Phenology Network (http://www.usanpn.org), will be essential for the understanding of (and eventually consideration of possible adaptations to) the coming impacts of climate warming on temperate plant communities.
We are grateful for advice and comments from I. Chuine and J. R. Strickler.