copan:LPJmL: a new hybrid modelling framework for dynamic land use and agricultural management

Abstract. Dynamic Global Vegetation Models (DGVMs) are established in environmental and agricultural sciences for many purposes, e.g., modelling plant growth and productivity, water and carbon cycles, and biosphere-climate interactions. Nevertheless, DGVMs are still rather limited in terms of simulating mutual interactions between biospheric and human processes. While DGVMs such as the Lund Potsdam Jena managed Land (LPJmL) model have been successfully connected to Integrated Assessment Models (IAMs), the model couplings often remain loose and static over the simulation period. The copan:LPJmL modelling framework is an extension of the copan:CORE framework for integrated and dynamic human-Earth system modelling, and addresses this issue by integrating LPJmL via a new interface, consisting of an LPJmL coupling library and a Python library pycoupler, which together enable LPJmL inputs and outputs to be coupled in copan:LPJmL during the simulation period. It uses the copan:CORE entities and integrates the coupled data into the World (simulation space as a whole) and the (grid) Cell entity, allowing other entities such as Individuals, e.g., for agent-based modelling (ABM), to access them. Besides ABM, this framework allows for a broad range of modelling approaches to be represented with copan:LPJmL, of which we introduce three examples: (1) The model of Integrated Social-Ecological Resilient Land Systems (InSEEDS), which uses a classical ABM approach to model management decisions by farmers, (2) an adaption of an established crop calendar model (Crop Calendar), and (3) a novel Large Language Model (LLM)-driven ABM approach (LLM Fertilization).