Advancing general ecosystem models (GEMs): Towards a mechanistic understanding of the biosphere in the light of the Anthropocene
Abstract
General ecosystem models (GEMs) are process‐based simulations of ecosystems functioning worldwide. They represent all essential processes in an ecosystem, going beyond a focus on specific biomes or organismal groups. GEMs are a comparably young development, and currently, there is still only one GEM available: the Madingley model, which was first announced in 2013. GEMs allow studying ecosystem mechanisms that are difficult or impossible to investigate empirically, and to perform simulation experiments to assess impacts of human pressures on ecosystems. As such, GEMs can be seen as the nature equivalent to ‘climate models'. As managers and policymakers urgently need good information on nature values and human impacts on them for policy design and decision‐making, they have much to gain from GEMs, which allow understanding current and predicting future outcomes of human actions on biodiversity and ecosystem functioning. To increase the applicability of GEMs for informing management and policy, we argue for an acceleration and diversification in model development, and we suggest a range of developments to propel GEMs forward not only towards both improved global representativeness but also towards applicability for more landscape‐level problems, a scale at which actual decision‐making often takes place. In addition, we argue for more accessible implementations and better documentation to increase both the transparency and credibility of GEMs, as well as the user basis applying GEMs. Practical implication. While some modelling developments target the global level, others are computationally only feasible for landscape‐level applications. We suggest that the development of GEMs should follow design principles employed in climate science and be structured so that they are modular and compatible in a standardized workflow.
What the paper shows and why it matters
(AI-generated)
Climate models, through simulation experiments and projections, have contributed significantly to climate policy. General ecosystem models (GEMs) were proposed by Purves et al. (2013) as the biodiversity counterpart. The paper uses the term narrowly, for process-based simulations of all trophic groups (levels of the food web) in any ecosystem on Earth. Managers and policymakers need this kind of mechanistic prediction of human impacts, but the field is young. Only one model fits the paper’s definition, the Madingley model (Harfoot et al. 2014), with a second still in development. It was designed for the whole globe at coarse resolution, whereas many management decisions are made at landscape scale. The paper asks how GEMs should be developed so that they can inform management and policy at both scales.
The paper argues that general ecosystem models should be developed faster and in several complementary directions, not as one model that does everything. It calls for global models alongside more detailed landscape-level models with the same goal of predicting biomass, flows of matter and body-size distributions. These should be modular, compatible and open-source, with shared workflows and common ways of handling uncertainty, as in climate science. The paper uses Madingley as its worked example. To show how such a model could support management and policy, it answers the three questions that Grimm et al. (2020) pose for models meant to support decisions (what the model is for, how it is organised, whether there is evidence it works), and it illustrates the model with a simplified simulation of South America, in which human use of plant production lowers the biomass of consumers. The gaps it names in Madingley include weak performance in strongly seasonal ecosystems, very simple vegetation, coarse animal movement, missing exchange between land and sea, and implementations that are hard to use without programming experience.
The paper is a research agenda, not a report of new results. It holds that GEMs are already useful for exploring hypotheses, for instance for gaining insight into the natural state of ecosystems that have no undisturbed counterpart. Active development, it argues, would make them relevant to research, management and policy from global to local scales, and would show where ecological knowledge is missing. The paper therefore judges investment in GEM development to be a high-gain, low-risk venture.
References
- Purves et al. (2013). Time to model all life on Earth. Nature 493, 295–297. doi:10.1038/493295a
- Harfoot et al. (2014). Emergent Global Patterns of Ecosystem Structure and Function from a Mechanistic General Ecosystem Model. PLoS Biology 12, e1001841. doi:10.1371/journal.pbio.1001841
- Grimm et al. (2020). Three questions to ask before using model outputs for decision support. Nature Communications 11, 4959. doi:10.1038/s41467-020-17785-2