Skip to main content

Concordance of long-term shifts with climate warming varies among phenological events and herbaceous species

Cite this dataset

Augspurger, Carol; Zaya, David (2020). Concordance of long-term shifts with climate warming varies among phenological events and herbaceous species [Dataset]. Dryad.


Many temperate herbs now flower earlier than a few decades ago. Little is known about other phenological events, despite the importance of life history integration for plant fitness. This study addresses the hypothesis that temporal shifts of multiple phenological events in herbs are concordant with temporal changes in weather. Explicitly showing that changes in timing of annual life cycle events are correlated with changes in weather-predicting variables provides support for the hypothesis that a phenological shift is concordant with climate change. Observations of six phenological events and five phenophases were made year-round for 25 yrs for herb species in a deciduous forest fragment, Trelease Woods in Illinois, USA. Dates for 43 species were analyzed by linear mixed-effects models for events and phenophases, and were compared to weather data from a nearby station. For early species, Emergence was delayed by 1.5 d/decade, while End Expansion advanced by 3.8 d/decade and Begin Dormancy advanced by 2.5 d/decade. For late species, End Expansion advanced by 6.7 d/decade, while Begin Senescence delayed by 17.7 d/decade. Begin Flowering and End Flowering advanced similarly for both seasonal groups, at 3.8 to 4.2 d/decade. Some events showed no temporal change. Species differed greatly in the degree or direction of change, related to seasonality of event or length of phase. Overall, for a given species, most events are advancing (68.4%) and most durations are shortening (74.4%). In 12 of 13 cases, inter-annual variation in event date was predicted by a temperature event-season combination variable, but in only six cases did both event date shift and weather variable warm through time. This finding supports the hypothesis that climate change is associated in changes in some, but not all, phenological events. This first long-term, multi-phase study of a community of temperate herb species indicates little temporal coherence of responses of multiple phases. Changes in date are event-specific, phase-specific, and species-specific. This complexity of responses among species and uneven responses within a species’ integrated annual cycle events has implications for evolutionary responses and more immediate interactions among plant, animal, and microbe species in this community.


DATA SET 1: 1993-2017

Observations of annual phenological events for 50 herbaceous species were made in Trelease Woods, a 24.5 ha fragment of temperate deciduous forest, dominated by canopy trees of sugar maple near Urbana, IL, USA.  Elevation varies by < 5m.  The 25 1-m2 quadrats observed were 50 m from the edge in a 50 m x 50 m grid in the north half of the Woods. 

The phenological status of each species in each quadrat was determined by the same observer (CKA) weekly in spring and summer, biweekly in fall, and biweekly to monthly in winter from 1993 – 2017.  For each species in each quadrat, dates of six phenological events were used to calculate five durations of phenophases.  Thus, measurements were recorded, and later analyzed, at the quadrat level.  Quadrats differed in number and size of individuals; regardless, the first (or last) individual in the quadrat to begin or end a phase defined the event date for that species in that quadrat

Species were analyzed together in seasonal groups (Appendix S2: Table S1).  It was not logistically practical to build climate models for 43 species separately for up to six events and five durations.  For most events, species were grouped as early or late (Table 1, Appendix S2: Table S1).

            We used linear mixed-effects models to test for phenological changes through time, and the weather variables that had the most predictive power of phenological events.  To address temporal trends in phenology, we tested 13 total phenological response variables in separate analyses for the dates of six events and 11 for the durations of five phases.  Individual quadrat observations, in each year, were used in the analyses.  For each phenological variable, we used model selection to find the top model at the seasonal group-level with respect to fixed effects and random effects structure.  All models included random effects terms for the species and quadrat, and the two were treated as crossed random effects.  Therefore, the analysis also provided species-specific temporal changes.

           For analyses of the relationship between phenological events and weather variables (Appendix SI: Text S3), we tested the dates of six phenological events in 13 separate analyses based on seasonal groups of species.  Fixed effects consisted of weather variables, and in some cases phenological variables that preceded the response, the “legacy” effects.  Weather variables described temperature and precipitation in individual months or seasons, and differed for each group of response variables (Appendix S2: Table S3).  The random effect of species was included in all models. 

            We completed three main steps to select weather variables that were most predictive of an event (see Appendix S1: Text S3 for details): 1) For each weather variable, individually determine the random and fixed effects structures for the top model predicting the phenological event.  Eliminate from further consideration those weather variables that were not included as a fixed effect. 2) Among correlated weather variables, choose the strongest predictor.  Exclude weather variables not in the top model predicting the phenological event.  Eliminate from further consideration those weather variables not included as a fixed effect, i.e. that are correlated with a stronger predictor.  3) Test different combinations of the remaining fixed effects, highlighting the top models and the weather variables they include.

            DATA SET 2: 1949-1964

            Observations of the same six events of eight of the same species (genus in case of Viola spp.) were made at unknown frequency from 1949-1964 by Charles Smith, woods custodian.  The location from 1949-1956 was noted as Brownfield Woods (5 yrs) or Trelease Woods (6 yrs) or both in the same year (3 yrs); no location was given for 1957-64.  Prior to 19th century deforestation, the two woods were in the Big Grove.  Today, the distance between the two fragments is <2 km.  Descriptions by Smith are: first appearance above ground (= Emergence by CKA), leaves full grown (= End Expansion), begin flowers (= Begin Flowering), end flowers (= End Flowering), leaves begin to color (= Senescence), leaves all fallen/withered (= Dormancy).  Differences between means for each of 44 species-events were determined by t-tests.

Usage notes

DATA SET 1: 1993-2017

A cell mumber for each of the six phenological events is the Julian calendar date in one year of the first (or last) individual of a species in one of 25 1 m x 1 m quadrats exhibiting the event.  Negative numbers for Emergence and End Expansion refer to number of days prior to Julian calendar date 1.  A cell number for each of the five phenological durations is the number of days derived from subtraction of dates of two events (e.g. Expansion - Emergence = Expansion Duration).

  Only non-italicized numbers were used in analyses (see Appendix S2: SI Table 5 for summary of species and phenological events used in analyses).  Species varied from 1 - 25 quadrats and 6 - 25 years of observation.  Primary criteria for a species to be analyzed were >15 years distributed relatively evenly over the 25 years; a minimum or constancy in number of quadrats were not required.

 Italicized numbers were not used in analyses, either because of too few years and/or additional criteria based on a species’ natural history, including complex phenologies such as multiple leafing periods, and ability to apply consistent criteria of an event (Appendix S1: Text S1), e.g. if an inter-annual pattern in plant size was obvious, the species was not analyzed.  

Missing data arose because a species appeared irregularly from year to year in a quadrat, an event did not occur in a quadrat in a given year (e.g. flowering, obvious senesence), or an event was not observed (particularly true of senescence/dormancy in autumn/winter or emergence of winter annuals in autumn).  The dominant reason for missing data or italicized data is provided under species name in the second column (see Appendix S1: Text 1 for details).  Also included are species nomenclature (following USDA, NRCS. 2019), family, and phenological syndrome, as well as miscellaneous notes of atypical phenology.  

DATA SET 2: 1949-1964

For 1949-1964, no methods are available, only data sheets.  A number is the Julian calendar date for a species in one of six phenological events in a given year; see Appendix S1: Text 1 for description of each event.  One date per species per year was recorded for each event; no variation was noted.  That date was assumed to represent mean values of a given event.  For comparison, mean values for each event for each year for each of the same eight species for 1993-2017 were calculated.