Testing a mechanistic general model of global ecosystems: improving prediction by increasing simplicity?

The Madingley Model (https://madingley.github.io/) is the first mechanistic general ecosystem model of ecosystem function and structure that is both global and applies to marine and terrestrial environments. It starts from microscopic events (births, deaths) and processes (metabolic rates) at the scale of the individual and scales up to dynamic and patterns at the macroscopic (ecosystem) scale. The current model makes a number of predictions that seem to match empirical patterns, eg inverted biomass pyramid for marine ecosystems; body size -growth rates of individuals, but others (eg heterotroph mortality) appear to fit less well.

The Madingley Model was conceived as a tool to simulate (predict) the impact of major disturbances (eg climate change, biological invasions) on biodiversity and ecosystem function. It is therefore important that the model is able to reproduce the patterns we observe in natural systems, but also that it does so with the minimum information required. This PhD will investigate whether there are functionally redundant processes and patterns in the current model, by analysing the sensitivity of the results (ecological patterns) from the current model to removal of components that relate to different biological processes. Subsequently, processes such as intelligent animal behaviour will be incorporated to consider whether there are key biological processes that are currently missing. Analyses will be based on computer simulations but there is also potential for mathematical analyses of sub-models in isolation.

The project will be based at the Centre for Biodiversity and Environment Research (https://www.ucl.ac.uk/cber), University College London, but the student will be co-supervised by Dr Drew Purves at Google DeepMind (http://drewpurves.com/), and Professor Dame Georgina Mace (CBER, UCL, https://www.ucl.ac.uk/cber/mace).

In addition to a passion for understanding ecological and ecosystem dynamics, as well as attention to detail, the successful candidate will be expected to have some or all of the following skills: computer programming (especially C/C++ or Java); mathematical modelling; statistical modelling. Applicants from a mathematics, engineering, computer science and/or physics background are particularly encouraged to apply.