Can we detect a nonlinear response to temperature in European plant phenology?

• Published in: International journal of biometeorology Vol. 60; no. 10; pp. 1551 - 1561
• Main Authors: Jochner, Susanne, Sparks, Tim H., Laube, Julia, Menzel, Annette
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.10.2016; Springer Nature B.V
• Subjects: Animal Physiology; bioclimatology; Biological and Medical Physics; Biophysics; cold; data collection; Earth and Environmental Science; ecosystems; Environment; Environmental changes; Environmental Health; Europe; Fagus; flowering; Flowers & plants; Flowers - growth & development; Leaves; Linear Models; Magnoliopsida - growth & development; Meteorology; Nonlinear Dynamics; nonlinear models; Nonlinear systems; Original Paper; Phenology; Picea - growth & development; Plant Leaves - growth & development; Plant Physiology; Seasons; spring; Temperature; Temperature effects; time series analysis; uncertainty; Europe; Climate change; Phenology; Europe; Nonlinearity; PEP725; Sigmoid; Temperature response
• DOI: 10.1007/s00484-016-1146-7
• ISSN: 0020-7128; 1432-1254

Over a large temperature range, the statistical association between spring phenology and temperature is often regarded and treated as a linear function. There are suggestions that a sigmoidal relationship with definite upper and lower limits to leaf unfolding and flowering onset dates might be more realistic. We utilised European plant phenological records provided by the European phenology database PEP725 and gridded monthly mean temperature data for 1951–2012 calculated from the ENSEMBLES data set E-OBS (version 7.0). We analysed 568,456 observations of ten spring flowering or leafing phenophases derived from 3657 stations in 22 European countries in order to detect possible nonlinear responses to temperature. Linear response rates averaged for all stations ranged between −7.7 (flowering of hazel) and −2.7 days °C −1 (leaf unfolding of beech and oak). A lower sensitivity at the cooler end of the temperature range was detected for most phenophases. However, a similar lower sensitivity at the warmer end was not that evident. For only ∼14 % of the station time series (where a comparison between linear and nonlinear model was possible), nonlinear models described the relationship significantly better than linear models. Although in most cases simple linear models might be still sufficient to predict future changes, this linear relationship between phenology and temperature might not be appropriate when incorporating phenological data of very cold (and possibly very warm) environments. For these cases, extrapolations on the basis of linear models would introduce uncertainty in expected ecosystem changes.