GEB: A large-scale agent-based socio-hydrological model – simulating 10 million individual farming households in a fully distributed hydrological model

Humans play a key role in the hydrological system, and their decisions influence the entire water system from tributary to river mouth. To fully comprehend how the human-natural water system evolves over space and time, and to investigate the systemic effects of climate change and human interventions, it is important to consider human behaviour and feedbacks to the hydrological system simultaneously at the local household- and large basin scales.

Therefore, we present GEB (Geographical, Environmental and Behavioural model); an agent-based model coupled to a fully distributed hydrological model that can simulate the behaviour and daily bi-directional interaction of more than 10 million individual farm households and reservoir operators with the hydrological system. Through this coupling, each individual farmer with unique characteristics and location can make daily decisions, such as irrigating their crops from surface-, reservoir-, or groundwater, planting and harvesting crops, investing in adaptation options (e.g., irrigation wells and sprinkler irrigation). All these decisions can be based on the available water in their environment, the status of their crops, their risk perception, crop price, water price, and weather conditions etc. Similarly, reservoir operators can regulate the availability of water for irrigation, and downstream releases of water based on the state of the hydrological system as well as communication with farmer agents.

GEB is dynamically linked with the spatially distributed hydrological model CWatM at 30’’ grid resolution (< 1km at the equator). Because many small-holder crop fields are much smaller, CWatM was specifically adapted to implement dynamically sized hydrological response units at field scale / sub-grid level, providing each agent with an independently operated hydrological environment.

While the model could be applied anywhere, we show an implementation with local and basin-wide feedbacks in the heavily managed Krishna basin in India, encompassing ~8% of India’s land area and ~12.1 million farmers. Here, we quantify bi-directional feedbacks such as the reservoir paradox and test various policies, such as providing subsidies for adaptation options (e.g., irrigation wells, sprinkler irrigation), and quantify effects on the hydrological system as well as downstream farmers.

In this implementation, GEB uses approximately 15 GB of RAM memory and can thus be used on an above average personal laptop. Computational requirements scale linearly with basin size, assuming similar farm-size distribution.