The need for unrealistic experiments in global change biology


  • Most multidriver temperature×CO2 phytoplankton experiments use only two levels of each driver.
  • Current studies cannot produce interaction surfaces for CO2 and temperature due to undersampling driver levels.
  • Categorisations of temperature×CO2 interactions are sensitive to small errors and biological variation.
  • Regression designs make more robust estimates of multidriver interactions than analysis of variance (ANOVA) driven designs.


Climate change is an existential threat, and our ability to conduct experiments on how organisms will respond to it is limited by logistics and resources, making it vital that experiments be maximally useful. The majority of experiments on phytoplankton responses to warming and CO2 use only two levels of each driver. However, to project the characters of future populations, we need a mechanistic and generalisable explanation for how phytoplankton respond to concurrent changes in temperature and CO2. This requires experiments with more driver levels, to produce response surfaces that can aid in the development of predictive models. We recommend prioritising experiments or programmes that produce such response surfaces on multiple scales for phytoplankton.

Collins S., Whittaker H. & Thomas M.K., 2022. The need for unrealistic experiments in global change biology. Current Opinion in Microbiology 68: 102151. doi: 10.1016/j.mib.2022.102151. Article.

  • Reset


OA-ICC Highlights

%d bloggers like this: