Dear Colleague,

Welcome to the first newsletter from the Behavioural Models of Land Systems (BeModeLS) working group. You might be thinking that you’ve already received newsletters from this working group, but those were likely from the group’s previous incarnation as the Large-scale Behavioural Models of Land Use Change working group. A small change in name accompanies some subtle shifts in emphasis to encompass a greater range of scales of model application and a more general consideration of land systems (beyond land use alone). The central theme remains the representation of human behaviour and social institutions in simulation models and we of course continue to be interested in the development of large-scale models of land use change. But we now also seek to support the creation of the next generation of land systems models that represent diverse human behaviour, agency, decision-making and institutional processes across scales and topics of application. 

We want this to be an inclusive, active, diverse and cooperative community of social-ecological land system modellers. As such, we encourage you to join our first online workshop in February 2023 to contribute to refining the proposed aims, objectives and future activities of the working group (more details below). You can see our current stated aims and objectives on the working group webpage, and more details about the workshop are below. Don’t hesitate to get in touch with the working group coordinator, James Millington, or members of the steering committee with your ideas for activities you would like to see the working group support - particularly if you’re unable to join the workshop. 

In this newsletter we also highlight learning resources, events and latest research that may be of interest to BeModeLS members, whether new and just establishing their modelling foundations or more established veterans with deep roots in the discipline. If you have something you’d like to highlight in a future newsletter, do let us know!   

Thanks for your continued support,
James Millington (@jamesmillington)
Calum Brown (@calum_brown1)
Tatiana Filatova (@TanjaFilatova)
Birgit Müller
Derek Robinson (@derektrobinson)
Mark Rounsevell
Maja Schlüter

In a previous working group workshop we launched a special issue on large-scale behavioural models of land use change in SESMO, the international, community-driven and open-access journal on Socio-Environmental Systems Modelling. Two papers have now been published in the special issue. Dressler et al. examine current challenges and promising strategies for upscaling in socio-environmental systems modelling, presenting an approach that facilitates the transfer of existing upscaling methods in ecology to socio-ecological systems. Perkins et al. report their initial steps towards developing a global behavioural model of anthropogenic fire for integration with an Earth system model, presenting results and analysis that show good model performance relative to external empirical measures of Human Appropriation of Net Primary Productivity (HANPP). Look out for more papers in the special issue coming soon.  

In a separate example of large-scale agent-based modelling, Brown et al. develop a new model of the British land system and apply it to detailed socio-economic (SSP) and climatic (RCP) scenarios. Using global boundary conditions from LandSyMM, they generate projections of land use and ecosystem service provision that account for a range of driving factors including individual and social behaviours, finding large areas of scenario space that can be explored with agent-based models of this kind. The model, CRAFTY-GB, is open-access and supported by a TRACE model evaluation protocol.

The topic of model evaluation and uncertainty is one that continues to be of importance for behavioural modellers, especially with respect to model parameters and structure. Two recent papers have contributed insights that may be of interest to practicing agent-based modellers. McCulloch et al. illustrate a process for calibrating an agent-based model’s structure and parameters that quantifies uncertainties through the combined use of History Matching and Approximate Bayesian Computation (admirably making all code and results available online). Troost et al. present a protocol to help reveal the often hidden premises for the choices made during model construction, ensuring reviewers and the broader modelling community can better assess the consistency and adequacy of those choices. This goal stems from the understanding that model validation should not be seen as an isolated step in the model process, and rather comes from demonstrating that choices in all steps of model construction (including parameter inference, uncertainty analysis, etc.) are adequate relative to the model’s intended use. Even for those not using agent-based approaches, both papers offer important ideas for evaluation of behavioural models. 

We welcome news, events, publications and job postings from the BeModeLS community. The deadline for content to appear in our next issue is 10 March 2023. Please submit items for consideration to James Millington.