Electronic Supporting Material: “Sensitivity of spring phenology to warming across temporal and spatial climate gradients in two independent databases”
1. Methods for NECTAR Database
The Network of Ecological and Climatological Timings Across Regions (NECTAR) database was designed to bring together species-rich plant phenological records with a focus on non-clonal, non-cultivated species observations of wild communities. Towards this goal, in October 2009, we reviewed the literature for ground (not herbarium) observational studies of species phenology, searching ISI for topic: (plant phenolog*) and Topic: (observat*) and Topic: (time-series OR long-term). Combined with personal knowledge, we identified 61 datasets. Of these, 12 datasets met all the conditions for our analysis here: (1) phenological data with ≥8 years of observations and (2) close proximity to daily temperature and precipitation data from the Global Historical Climatology Network (GHCN). When possible, we worked with data holders or those very familiar with the data to guide quality control, merging, and taxonomy issues.
These 12 datasets covered a range of habitats and represented a mix of citizen science observations and various initiatives by scientific organizations (see list below). Because datasets were collected with a variety of objectives and methods, we took special care to merge them into one database. The most common forms of data were first-leaf or first-flower dates (FLD and FFD), and we adjusted more highly resolved data to these formats and excluded phenological events, such as fruiting, for which we had little cross-site data. For sites with general monitoring data (where species were regularly observed as ‘in-flower’ and so on), we took FLD and FFD as the first occurrence of leaf or flower for each species each year. For GTH, LUQ, and SEV, we first aggregated by plot level and then took FFD and FLD for each species in each year. For our site with monitoring data and nearly year-round flowering (LUQ), we consulted with scientists familiar with the site to set the start of year date for selecting FLD and FFD (January 1). For all datasets we removed any species noted as cultivated in any way. After aggregating to FLD and FFD by species and year we merged the 12 datasets using R version 2.12 ( (R Development Core Team 2010).
Because the 12 NECTAR data sets represented a mix of observers, timeframes and locations, we controlled for variation in taxonomic identification as much as possible. We compared all names to a list from the International Plant Names Index ( and, for any species that did not match, we checked spelling and established synonyms and corrected names when possible. In a small number of cases correcting to a well-established synonym led to duplicate observations of a given species at a given site within a year. In such cases we took the first observation.
We review basic information on each dataset below, referring readers to the relevant publications, metadata, and (where possible) raw data for more detailed information. Also included are the location and identifier numbers for the associated GHCN station from which daily climate data were taken. Where applicable, URL links have been embedded to provide interested parties with direct access to the data or data providers. As mentioned in the main text, GHCN station selection was based on both proximity to the site of the phenology data and the temporal overlap between the phenological data and climate data.
2. Description of NECTAR datasets
•CHN– First-flowering dates for 384 species in and around (generally within several kilometers, all observations within 80 kilometers) Chinnor, Oxfordshire,England (GRSU7599) from 1954-2000, all recorded by R.S.R. Fitter (Fitter and Fitter 2002; Fitter and others2005). The complete dataset was published as supplemental material by Fitter and Fitter (2002). Climate Data from GHCN Station UK000056225; 51.77oN, 1.27oE
•FAR – First-flowering dates for 675 species collected from the Great Plains (these data cover North Dakota and Minnesota, USA). Data from 1910-1961 were collected by a single observer (Orin A. Stevens) and are publicly available through Ecological Archives (Dunnelland Travers 2011; Travers and others2009). NECTAR also includes data from 2007-2010 collected by S. E. Travers and colleagues. Climate Data from GHCN Station USC00212142; 46.84oN, 95.84oW
•GTH– Every other day flower monitoring for 116 species (Inouye 2008; Lambert and others2010; Miller-Rushing and Inouye 2009) from a series of plots located at the Rocky Mountain Biological Station (Gothic, Colorado) from 1973 to present (NECTAR includes data up to 2009, but data collection annually is active). Data were graciously provided by the data collector, David Inouye, who has monitored, or supervised monitoring, over the past 37 years. Support for collection of these data came from NSF grant DEB 0922080. Climate Data from GHCN Station USC00051959; 38.87oN, 106.98oW.
•HUB – Weekly observations of leaf phenological events of 3 tree plant species from 1989-2008 at Hubbard Brook Ecosystem Study, New Hampshire (USA). Data collection headed by Amey Bailey. Data are publicly available ( following Long Term Ecological Research nework (LTER) policies. Data were provided by the Long Term Ecological Research (LTER) Program. Climate Data from GHCN Station USC00275995; 44.03oN, 71.14oW
•HVD– Weekly observations (during the early and late season, April-June and September leaf-fall) of leaf phenological events of 33 woody plant species from 1990-2009 (with fewer species recorded for the entire observation period) at Harvard Forest, Massachusetts (USA). We consider first-leaf date to be the first observation for each species for each year when the percentage of buds broken open was greater than zero (BBRK >0) and first-flower date as the first observation when the percentage of flower buds open was greater than zero (FOPN >0). Data are publicly available ( following LTER policies. Data were provided by the Long Term Ecological Research (LTER) Program. Climate Data from GHCN Station USW00094746; 42.27oN, 71.87oW
•KON– First-flowering dates for 239 species recorded from Konza Prairie LTER from 2001-2009 by volunteers ( Monitoring organized by the Konza Environmental Education Program (KEEP). Climate Data from GHCN Station USC00144972; 39.20oN, 96.58oW
•LUQ– Litter trap collections of fallen leaves, flowers, fruit and seeds from 86 species from Luquillo Experimental Forest Long Term Ecological Research Network in Puerto Rico (Zimmerman and others2007). Baskets were arrayed in two forests: El Verde (1992-2000 and 2006-2007 data) and Bisley (1992-1994). Data are publicly available ( following LTER policies. Special thanks to Jess Zimmerman, Chris Nytch, JimenaForero-Montaña for providing the data. Collection of these data was supported by NSF DEB grants: #9411973, #0080538, #0218039, #0620910, #0614659, #0218039. Climate Data from GHCN Station RQW00011641; 18.43oN, 66.01oW
•MHK–Observations(1928-2002)offirstfloweringdatefor18perennialspeciesatMohonkLakeResort and Preserve, outside of New Paltz, New York, USA (Cook and others2008). Observations were taken along pre-determined routes at daily intervals during the spring to minimize observer biases. Climate Data from GHCN Station USC00305426; 41.77oN, 74.15oW
•SEV– Data sets were provided by the Sevilleta Long Term Ecological Research (LTER) Program. Significant funding for the collection of these data was provided by the National Science Foundation Long Term Ecological Research program (NSF Grant numbers BSR 88-11906, DEB 9411976, DEB 0080529 and DEB 0217774). Data for this analysis were drawn from a combination of two datasets of 182 species recorded from 1991-1994 (‘Phenology transects’) and 2000- 2008 (‘Core site phenology’) including a variety of phenological events. Core site phenological data were collected every month on four transects and included data on the first 10 individuals of each species encountered. Transects were collected weekly during June-August and at least monthly for the rest of the year from 3-4 transects each in four different habitat types. Data are publicly available following LTER policies (available through metacat/knb-lter-sev.13700.1, metacat/knb-lter-sev.4895.1, metacat/knb-lter-sev.4894.1, metacat/knb-lter-sev.4893.1, metacat/knb-lter-sev.48.1, metacat/knb-lter-sev.4892.1). Climate Data from GHCN Station USC00292436; 32.25oN, 107.75oW
•UWM– Weekly observations of leaf phenological events of 27 woody species from 2000-2009 at University of Wisconsin-Milwaukee Field Station, located in Saukville, Wisconsin (Schwartz and Hanes 2010). Methods closely followed those of Harvard Forest (USA). Climate Data from GHCN Station USC00473058; 43.24oN, 88.12oW
•WAU– First leaf and first bloom dates from Wauseon, Ohio recorded by Thomas Mikesell from 1883-1912 for 26 tree species (Smith 1915). These data have been used in a number of comparative phenological studies (Hunter andLechowicz 1992; Morin and others2009). Climate Data from GHCN Station USC00338822; 41.52oN, 84.15oW
•WDC– First-flowering dates recorded by over one hundred volunteers for 778 species in the Washinton, DC metropolitan area (USA) from 1985-2007 (Abu-Asaband others2001). Data collections were guided by A. Goldberg and J. Wurdack and organized through the Department of Botany, Smithsonian National Museum of National History with data available online ( The complete dataset is far greater and more species-rich than the one included in NECTAR—we reduced the data by removing species that were flagged as cultivated. Climate Data from GHCN Station USW00013721; 38.30oN, 76.42oW
3. PEP725 Data
Please see the main text for a full description of, and references for, the PEP725 database. Descriptions of the PEP725 data we used for this analysis, including species names and phenological phases, are in ESMTable 1 (for FFD related phases) and ESM Table 2 (For FLD related phases).
3. Sensitivity Tests
We conducted a suite of sensitivity tests to determine the robustness of our results to outliers, the standardization period chosen for the climate predictors, and the minimum sample size threshold that we used.
To test the sensitivity of our results to outliers, we repeated our main analysis of the NECTAR database, using two levels of outlier screening to remove observations that lay beyond 2-sigma or 3-sigma standard deviations from the mean. In general, our results in terms of the percentage of significant responders (ESM Table 3) and mean site sensitivity (ESM Table 4) are robust to the level of outlier screening we choose, including the default (no screening) in the manuscript. Largest site differences are found for SEV and LUQ using the more restrictive 2-sigma screening level. These two sites have the lowest proportion of significant temperature responders: as with the standardization period test (see below), the overall low number of significant responders at this site means any slight change in the significant responders will have a relatively large impact on the percentage of significant responders and the mean calculated sensitivity.Importantly, our cross-site regression (percentage of significant responders versus mean annual temperature) is also robust, irrespective of the outlier filtering used (ESM Table 5 and 6). Our full regression (All Sites) is still significant at the p<0.01 level, with similar explanatory power (R2), and a similar slope (beta coefficient). Results are also similar for the alternate regression, in which SEV and LUQ are removed. From these analyses, we conclude that our overall results are insensitive to choices we make regarding outliers.
To test the robustness of our results to the standardization periods used, we repeated our main analyses, but restricted the temporal range to the last 30 years of the 20th century (1970-1999). For this analysis, only 7 NECTAR sites had at least 8 observations during this interval: CHN, GTH, HUB, HVD, LUQ, MHK, and WDC (ESM Table 7).The temperature record for HUB begins in 1975, but at these other 6 sites the temperature data is 100% serially complete over the entire 1970-1999 interval. This allows us to use a standardization time frame for the GDDsum predictors common to all these sites (except HUB), analogous to the procedure used for standardizing thePEP725 climate predictors. The percentage of significant responders and the mean sensitivities were similar between this analysis and the original, where we used all the data and standardized each site using different time periods (ESM Tables 8 and 9). The greatest difference is in the mean sensitivity for LUQ: the overall low number of significant responders at this site means any slight change in the significant responders will have a large impact on the mean calculated sensitivity. From this additional analysis, we conclude that our results and conclusions are largely insensitive to the standardization period chosen for the GDDsumpredictors.
The primary reason we chose a minimum time series length of 8 observations was so that we could include several important sites that contain many short records (for example, Konza, Sevilleta, and University of Wisconsin; median time series lengths are contained in Table 1 of the main manuscript).To test the robustness of our results, we have now repeated our initial analysis, using longer minimum series lengths of 15 and 20 years. This more restrictive length criteria limits the NECTAR sites we can use: CHN, FAR, GTH, HUB, HVD, MHK, and WDC. And even for these sites, these new minimum length thresholds substantially reduced the number of time series available (ESM Table 10).Notably, the number of time series available at FAR, HVD, and WDC drop by approximately 40-60% with a requirement of n>=15. With the higher n criteria, the percentage of significant responders does tend to increase (ESM Table 11). The changes are relatively modest, however, and the cross-site ranking is generally retained: for example, of these sites, WDC still has the lowest proportion of significant temperature responders, whereasGTH still has the highest proportion. Notably, the sites with the largest changes are also the sites where the new length criteria resulted in the largest declines in available time series (FAR and WDC). It is unclear whether the increase in significant responders is due to increased power with these longer time series, or undersampling of species that are less sensitive to temperature.Mean sensitivities for the significant responders also generally increase, although, like the percentages, have a similar magnituderegardless of the minimum n chosen (ESM Table 12).
References
Abu-Asab MS, Peterson PM, ShetlerSG, Orli SS (2001) Earlier plant flowering in spring as a response to global warming in the Washington, DC, area. Biodiversity and Conservation, 10, 597–612.
Cook BI, Cook ER, Huth PC, Thompson JE, Forster A, Smiley D (2008) A cross-taxa phenological dataset from Mohonk Lake, NY and its relationship to climate. International Journal of Climatology, 28, 1369–1383.
Dunnell KL, Travers SE (2011) Shifts in the flowering phenology of the Northern Great Plains: patterns over 100 years. American Journal of Botany, 98, 935–945.
Fitter AH, Fitter RSR (2002) Rapid changes in flowering time in British plants.Science, 296, 1689.
Fitter AH, Gilligan CA, Hollingworth K, Kleczkowski A, TwymanRM, PitchfordJW (2005) Biodiversity and ecosystem function in soil. Functional Ecology, 19, 369–377.
Hunter AF, LechowiczMJ (1992) Predicting the timing of budburst in temperate trees. Journal ofApplied Ecology, pp. 597–604.
Inouye DW (2008) Effects of climate change on phenology, frost damage, and floral abundance ofmontane wildflowers. Ecology, 89, 353–362.
Lambert AM, Miller-Rushing AJ, Inouye DW (2010) Changes in snowmelt date and summer precipitation affect the flowering phenology of Erythronium grandiflorum(glacier lily; liliaceae). American Journal of Botany, 97, 1431.
Miller-Rushing AJ, Inouye DW (2009) Variation in the impact of climate change on flowering phenology and abundance: An examination of two pairs of closely related wildflower species. American Journal of Botany, 96, 1821.
Morin X, LechowiczMJ, Augspurger C, O’Keefe J, Viner D, Chuine I (2009) Leaf phenology in 22 North American tree species during the 21st century.Global Change Biology, 15, 961–975.
R Development Core Team (2010) R: A Language and Environment for Statistical Computing.R Foundation for Statistical Computing, Vienna, Austria.URL 3-900051-07-0.
Schwartz MD, Hanes JM (2010) Intercomparing multiple measures of the onset of spring in eastern North America. International Journal of Climatology, 30, 1614–1626.
Smith JW (1915) Phenological dates and meteorological data recorded by Thomas Mikesell at Wauseon, Fulton County, Ohio. Monthly Weather Review Supplement, 2, 21–93.
Travers SE, Dunnell KL, Michener WK (2009) First-flowering dates of plants in the Northern Great Plains. Ecology, 90, 2332.
Zimmerman JK, Wright SJ, Calderón O, Pagan MA, Paton S (2007) Flowering and fruiting phenologies of seasonal and aseasonal neotropical forests: the role of annual changes in irradiance. Journal of Tropical Ecology, 23, 231–251.
ESM Table 1: First Flowering Data from the PEP725Database Used in our Analysis, Encompassing 91,808Phenological Time Series, 50 Unique Events, and 42 Unique Taxa
Code / Species / Phase / # Series / Median n101060 / Aesculus hippocastanum / First flowers open / 4837 / 25
102060 / Alnusglutinosa / Beginning of flowering / first pollen spread / 3115 / 20
103060 / Alopecuruspratensis / Beginning of flowering / first pollen spread / 1603 / 13
105060 / Artemisia vulgaris / Beginning of flowering / first pollen spread / 1256 / 12
107060 / Corylusavellana / Beginning of flowering / first pollen spread / 4158 / 22
109060 / Forsythia suspensa / First flowers open / 3966 / 22
110060 / Picea abies (P.excelsa) / Beginning of flowering / first pollen spread / 1831 / 15
111060 / Quercus robur (Q.peduncula) / First flowers open / 19 / 24
112060 / Sambucus nigra / First flowers open / 4942 / 23
113060 / Dactylisglomerata / Beginning of flowering / first pollen spread / 109 / 23
114060 / Tussilago farfara / First flowers open / 4623 / 22
115060 / Acer platanoides / First flowers open / 3600 / 20
116060 / Acer pseudoplatanus / First flowers open / 1 / 42
117060 / Alnusincana / Beginning of flowering / first pollen spread / 22 / 22
118060 / Anemone nemorosa / First flowers open / 4286 / 23
119060 / Betula pubescens / First flowers open / 27 / 16
120060 / Fraxinus excelsior / First flowers open / 2886 / 18
121060 / Galanthusnivalis / First flowers open / 5124 / 25
123060 / Prunus spinosa / First flowers open / 3771 / 22
124060 / Robiniapseudoacaci / First flowers open / 3431 / 20
125060 / Salix caprea / First flowers open / 4744 / 25
126060 / Sorbusaucuparia / First flowers open / 3575 / 20
127060 / Syringa vulgaris / First flowers open / 4986 / 25
128060 / Taraxacum officinale / First flowers open / 4915 / 25
129060 / Tiliacordata / First flowers open / 2592 / 18
220060 / Malus x domestica (e. cultivar) / First flowers open / 1278 / 12
221060 / Malus x domestica (l. cultivar) / First flowers open / 1315 / 11
222060 / Prunus/Cerasus avium (e. cultivar) / First flowers open / 190 / 29
223060 / Prunus/Cerasus avium (l. cultivar) / First flowers open / 11 / 44
224060 / Vitis vinifera / First flowers open / 413 / 18
225060 / Prunus domestica / First flowers open / 29 / 30
227060 / Pyrus communis (e. cultivar) / First flowers open / 48 / 27.5
228060 / Pyrus communis (l. cultivar) / First flowers open / 8 / 44
229060 / Ribes rubrum / First flowers open / 4514 / 21
230060 / Malus x domestica (u. cultivar) / First flowers open / 119 / 23
232060 / Prunus/Cerasus avium (u. cultivar) / First flowers open / 145 / 20
235060 / Prunus domestica (u. cultivar) / First flowers open / 105 / 21
237060 / Pyrus communis (u. cultivar) / First flowers open / 96 / 23
332060 / Secalecereale (winter) / First flowers open / 2561 / 14
332061 / Secalecereale (winter) / Beginning of flowering - first anthers vis / 2546 / 13
337061 / Helianthus annuus / First flowering / 238 / 11
660060 / Calluna vulgaris / First flowers open / 3585 / 21
661060 / Cornus mas / First flowers open / 11 / 44
663060 / Fragariavesca / First flowers open / 72 / 13
666060 / Vacciniummyrtillus / First flowers open / 93 / 21
669060 / Populustremula / First flowers open / 3 / 12
780060 / Laurusnobilis / First flowers open / 2 / 27
781060 / Olea europea / First flowers open / 2 / 44.5
782060 / Prunus amygdalis/dulcis / First flowers open / 4 / 44
783060 / Rosmarinus officinalis / First flowers open / 1 / 44
Designators e./l./u. refer to early, late, and unknown cultivars, respectively. “Code” refers to the taxa/phase code in the PEP725 database, “Phase” is the descriptor given in the PEP725 database, “# Series” is the number of time series available for that taxa/phase code, and “Median n” is the median number of observations for all time series for that taxa/phase code.