Projects per year
Landscape fire regimes are created through socio-ecological processes, yet in current global models the representation of anthropogenic impacts on fire regimes is restricted to simplistic functions derived from coarse measures such as GDP and population density. As a result, fire-enabled dynamic global vegetation models (DGVMs) have limited ability to reproduce observed patterns of fire, and limited prognostic value. At the heart of this challenge is a failure to represent human agency and decision-making related to fire. This paper outlines progress towards a global behavioural model that captures the categorical differences in human fire use and management that arise from diverse land use objectives under varying socio-ecological contexts. We present a modelled global spatiotemporal distribution of what we term ‘land-fire systems’ (LFSs), a classification that combines land use systems and anthropogenic fire regimes. Our model simulates competition between LFSs with a novel bootstrapped classification tree approach that performs favourably against reference multinomial regressions. We evaluate model outputs with the human appropriation of net primary production (HANPP) framework and find good overall agreement. We discuss limitations to our methods, as well as remaining challenges to the integration of behavioural modelling in DGVMs and associated model-intercomparison protocols.