You are here
Overview of Scientific Findings and Applications Stemming from the Historic North American Lilac-Honeysuckle Observation Network
Prior to the 1980s, traditional simple models of phenological development were often constructed based on measuring degree-day accumulations from some arbitrary or calculated starting date until the time of the phenological event (stage) being studied. Several limitations are implicit in this approach: 1) degree-day accumulations are treated as a bucket model, in that no consideration is usually given to the manner in which they add up; only the time that the total passes a predetermined threshold is regarded as important; and 2) plant-climate relationships are deemed to be highly site-specific, such that regional or continental-scale phenological models are not practical.
While several earlier researchers (Wang 1960, Caprio 1974) had previously taken steps toward addressing these limitations, Schwartz and collaborators initiated a more comprehensive reevaluation of them in several studies during the mid-1980s (Schwartz 1985; Schwartz and Marotz 1986, 1988). Specifically, these papers demonstrated that the timing and magnitude of periods of warm air advection (driven by synoptic-scale weather systems) were a superior method of variable selection, and models were developed based on data from stations distributed over a large (3,000 km x 2,000 km) region in eastern North America. Further, Schwartz (1985) and Schwartz and Marotz (1986) developed and tested continental-scale phenological models based on cloned lilac (Syringa chinensis ‘Red Rothomagensis’) first leaf event data recorded in the eastern North America network, finding that such models could successfully predict phenological responses with an acceptable average absolute error of eight days or less. Eventually, these models were refined and calculations expanded to include an extensive range of measures, termed the “Spring Indices (SI) Suite of Measurements” (Schwartz 1997; Schwartz and Reiter 2000).
SI has allowed assessment of the general impact of climate change on the onset of spring in North America, China, and across temperate land areas of the Northern Hemisphere (Schwartz and Reiter 2000; Schwartz & Chen 2002; Schwartz et al. 2006). Cayan et al. (2001) established the connection of phenology to temperatures and stream runoff in the western USA. Current work is exploring the use of spring phenology as a predictive measure of summer fire severity in the western USA, given the impact of the timing of plant growth on soil water depletion rates (Westerling et al. in prep.).
In general, assessments of temporal change require standardized methodologies and tools. The 50-year lilac datasets and their resultant products are useful because of the consistency of methodology and tools applied. However, use of a particular tool through time, when resources are limited, requires the tradeoff of generality. The greatest values of using a clonal indicator plant (such as lilac) as the basis for a continental-scale phenology network are: 1) consistent response to environmental variation; 2) broad distributional range; 3) virtual certainty that the correct plant is observed; and 4) the provision of a standard reference that can be compared among different sites and years. Ideally, models based on clonal indicator plants will also be readily connected to sequences of weather events and changes in characteristics of the lower atmosphere. Any such variables proposed as “spring indices” would need to be thoroughly tested before adoption, and could involve a suite of related measures.
A “first approximation of a comprehensive spring index” was proposed based on an averaging of the modeled dates of first leaf models for the lilac and honeysuckle indicators (Schwartz 1990). Next, a first leaf model-derived “spring index” facilitated reconstructing past phenological responses in eastern North America (Schwartz 1993). The last spring frost (–2.2°C) date and a new measure, the “damage index” (spring index first leaf date minus last spring frost date) were calculated to assess the risk of frost damage to spring plant development. A “second generation” of models added the first bloom date calculations to the first leaf dates produced by the original versions (Schwartz 1997). The “Spring Indices (SI) suite of measurements” was introduced to evaluate 1900-1997 springtime changes in North America (Schwartz and Reiter 2000). The results showed that the SI first leaf date advanced by 5.4 days, the first bloom date by 4.2 days, and the last frost date by 4.2 days from 1959-1993, which compared favorably with a 6.0 day advance for the same period reported in Europe (Menzel and Fabian 1999).
Application of the SI models to temperate regions of the Northern Hemisphere indicated an advance of early spring warmth (SI first leaf date, -1.2 days/decade), late spring warmth (SI first bloom date, -1.0 days/decade), and last spring freeze date (-1.5 days/decade) from 1955 to 2002 (Schwartz et al. 2006). Unlike North America and Europe, China showed a weaker long-term trend toward earlier spring warmth from 1959-1993, but stronger advances in last spring freeze dates (Schwartz and Chen 2002).
When plants break dormancy and foliage reappears in spring, a rapid increase in transpiration and changes in surface albedo alter thermodynamic properties of the surface layer. Normally, seasonal variations in tropospheric characteristics and year-to-year variability in the green-up date (which can range up to 1.5 months) mask the impact of these effects on surface maximum temperatures. However, these air temperature changes were detected when phenology of cloned lilacs was used as the indicator of the onset of transpiration (Schwartz and Karl 1990). Moreover, the “green wave” (i.e., onset of empirically observed and modeled leaf-out) occurs in conjunction with discontinuities in lower atmospheric temperature, humidity, and wind (Schwartz 1992). These signals are all consistent with the start of plant photosynthetic activity, seasonal shifts in atmospheric circulation patterns, and physical changes in the nature of the surface layer (Schwartz 1992).
Fitzjarrald et al. (2001) developed an effective method for identifying the timing of this abrupt shift in lower atmospheric conditions, and the diurnal temperature range (DTR, difference between daily maximum and minimum air temperature) shows a discontinuity at the onset of spring that coincides with the onset of plant transpiration as measured by first leaf of cloned lilacs (Schwartz 1996). These results further defined the onset of spring in mid-latitudes as modally abrupt rather than as a gradual seasonal transition.
Changes in lower atmospheric characteristic during spring are ultimately related to variations in the surface energy balance, and have implications for carbon dioxide flux (e.g., net ecosystem exchange, NEE) and balance. In recent years, direct measurement of these variables at eddy-covariance sites in the Ameriflux (western hemisphere) and Fluxnet (worldwide) networks have made more extensive studies of phenological interactions possible. For example, leaf emergence was one of the important factors that controlled seasonal changes in latent and sensible heat fluxes at a single site (Wilson and Baldocchi 2000) and across several sites within the eastern deciduous forests of North America (Schwartz and Crawford 2001). These two studies indicate that relationships between phenology and surface energy balance are qualitatively similar across different land cover types and geographical locations, and that inter-site variation could be explained by differences in land cover. An ongoing extension of these studies will further confirm these geographical and land cover relationships (Schwartz and Hanes, in preparation).