Identifying the best climatic predictors in ecology and evolution. Issue 10 (11th July 2016)

• Record Type: Journal Article
• Authors: van de Pol, Martijn; Bailey, Liam D.; McLean, Nina; Rijsdijk, Laurie; Lawson, Callum R.; Brouwer, Lyanne
• Editors: Gimenez, Olivier
• Is Part Of: Methods in ecology and evolution. Volume 7:Issue 10(2016:Oct.)
• Page Start: 1246
• Page End: 1257
• Publication Date: 2016-07-11
• Subjects: bias -- climate change -- climate sensitivity -- cross‐validation -- false positive -- precision -- R package climwin -- sample size -- sliding window -- weather; Ecology -- Periodicals; Evolution -- Periodicals; 577
• DOI: 10.1111/2041-210X.12590
• Languages: English
• ISSNs: 2041-210X

Abstract

Summary: Ecologists and many evolutionary biologists relate the variation in physiological, behavioural, life‐history, demographic, population and community traits to the variation in weather, a key environmental driver. However, identifying which weather variables (e.g. rain, temperature, El Niño index), over which time period (e.g. recent weather, spring or year‐round weather) and in what ways (e.g. mean, threshold of temperature) they affect biological responses is by no means trivial, particularly when traits are expressed at different times among individuals. A literature review shows that a systematic approach for identifying weather signals is lacking and that the majority of studies select weather variables from a small number of competing hypotheses that are founded on unverified a priori assumptions. This is worrying because studies that investigate the nature of weather signals in detail suggest that signals can be complex. Using suboptimal or wrongly identified weather signals may lead to unreliable projections and management decisions. We propose a four‐step approach that allows for more rigorous identification and quantification of weather signals (or any other predictor variable for which data are available at high temporal resolution), easily implementable with our new R package 'c limwin '. We compare our approach with conventional approaches and provide worked examples. Although our more exploratory approach also has some drawbacks, such as the risk of overfitting and bias that our simulations show can occur at low sample and effect sizes, these issues can be addressed with the right knowledge and tools. By developing both the methods to fit critical weather windows to a wide range of biological responses and the tools to validate them and determine sample size requirements, our approach facilitates the exploration and quantification of the biological effects of weather in a rigorous, replicable and comparable way, while also providing a benchmark performance to compare other approaches to.