Figure 2.

Four different ways to determine the biological response function: using (a) temporal or (b) spatial variation in observational studies, (c) experimental manipulation and (d) mechanistic modelling. The examples also highlight the diversity in response variables, from (a) phenological and (c) developmental phenotypic traits to (d) population and (b) ecosystem parameters. (a) Observational study relating temporal variation in the timing of egg laying (days since April 1) to annual variation in spring temperatures using linear regression on 47 years of data on wild-living British chaffinches (Fringilla coelebs) [25]. (b) Observational study relating spatial variation in annual primary plant productivity to spatial variation in precipitation using linear regression on data from 11 ecosystems [26]. (c) Experimental study determining the thermal performance curve for daily growth rates of hornworm larvae (Manduca sexta) using five different levels of experimentally manipulated rearing temperatures in the laboratory [27]. (d) Mechanistic study using a population matrix model parameterized with temperature-dependent demographic rates to calculate how the population growth rate of Daphnia lumholtzi depends on temperature [28]. Note that in (c,d) the climatological distribution can be derived from climatological time series (similar to blue panel of a), but that determining the distribution of biological response requires additional observations, as simply imposing the climate distribution to the response function ignores other sources of variation in biological response.