GEB: a socio-hydrological model for risk management on a European scale

We present the Geographical, Environmental and Behavioural (GEB) socio-hydrological model at the pan-European scale. The model is agent-based, representing the interactions between (local) hydrology and human water management for hundreds of millions of people across Europe. For this, GEB couples an agent-based adaptation model, a fully distributed hydrological model (originally forked from CWatM), and a hydrodynamic model (SFINCS). Within GEB, people can dynamically respond to their environment, and the decisions that agents make also can affect the environment. For example, farmer agents can change their crops and adopt other measures such as wells in response to droughts, which in turn affects ground- and surface water in the hydrological model. Household agents can adapt to changes in flood risk and respond to flood events by wet- or dry-proofing their house. All adaptation decisions consider heterogeneity in the agent population, such as differences in age and education level. The model architecture allows for a fully automated setup of the model. To initialize the model, and to allow for parallel computing, river basins are automatically clustered based on size and proximity, after which those clusters are run in parallel using an efficient Snakemake workflow. The new hydrological model in GEB simulates hydrological fluxes on an hourly timestep. To validate the model, we compare simulated discharge with discharge observations from the Global Runoff Data Centre (GRDC) dataset, focussing on high flows during flood events. Subsequently, the skill scores (e.g. Kling–Gupta efficiency, KGE) are compared to state-of-the-art hydrological models such as LISFLOOD. GEB is currently used for a wide range of applications, such as (but not limited to) assessments of drought and flood risk, extreme weather impacts (e.g. hail), multi-risk, household adaptation measures, nature-based solutions and early warning. The model is open source and can be accessed via https://github.com/GEB-model/GEB.