Data-driven agent-based modelling of incentives for carbon sequestration: The case of sown biodiverse pastures in Portugal

Sown biodiverse permanent pastures rich in legumes (SBP) offset animal farming emissions due to their potential to sequester carbon. From 2009 to 2014 Portugal implemented a programme that provided payments to incentivize the adoption of SBP. However, no proper evaluation of its outcome was conducted. To address this gap, we develop an agent-based model (ABM) at the municipality level to study the adoption of SBP in Portugal and assess the outcome of the programme. We applied the first pure data-driven approach in agricultural land-use ABM, which relies on machine learning algorithms to define the agents' behavioural rules and capture their interaction with biophysical conditions. The ABM confirms that the program effectively expanded the adoption of SBP. However, our estimates indicate that the adoption rate in the absence of payments would have been higher than originally predicted. Furthermore, the existence of the program decreased the adoption rate after its conclusion. These findings underscore the importance of using reliable models and considering residual effects to properly design land use policies. The ABM developed in this study provides a basis for future research aimed at supporting the development of new policies to further promote the adoption of SBP.