The model architecture described in this deliverable provides the framework through which data and simulations from the data on food waste at a consumer level and at a retail level can be integrated into simulation models. This report highlights the technical approaches followed to achieve model integration.
This report highlights the technical approaches followed to achieve model integration. An integrated whole-of-system modelling approach will be developed as a part of the REFRESH project to allow the development of a decision-relevant, and dynamic policy support tool, by which a road map to the reduction of European food waste by 50% by 2030 can be developed. The vital first step (highlighted in this report) is to develop the structures to allow model integration between different model types: Agent-Based Models and Bayesian Networks. These structures were developed and tested to ensure that the model types can be integrated. The architecture described in this deliverable provides the framework through which data and simulations from the data on food waste at a consumer level and at a retail level can be integrated into simulation models.
Since a sizable share of the food waste is generated either at the consumer level or at the interaction between consumers and retailers, we address the modelling effort with two integrated ABM-BN models. The first model reproduces the dynamic evolution of food waste choices of consumers as consequence of social interactions. The second focuses instead on the conditions for the successful diffusion and adoption of innovations to reduce food waste at the retailer level.
The systemic modelling approach proposed will allow the development of selected simulation scenarios at the consumer and retail level, facilitating decision making in the face of uncertainty.
These integrated setups are first iterations of working integrated models, aimed at validating technically the setups as well as the integration process itself. As they are, there are certainly factors that are likely to be important in determining food waste, which are not yet included in the models. However, the latter are flexible and can accommodate further details, and variables. Their construction is purposefully flexible in terms of components of decisions. The integration with Bayesian Networks ensure that Agent-Based models will learn from data originated from the other refresh WPs and will evolve, allowing the introduction of new variables and factors that will lead to the improvement of the different simulation scenarios.
The REFRESH project implements a behavioural economics approach in order to identify and measure the most important socio-economic conditions and potential policy interventions driving businesses’ and consumers’ choices in the generation of food waste. More specifically, this work aims to provide new information on consumer and business behaviour by measuring the effects of major tangible factors of food waste, by identifying hidden and emerging profiles of consumer’ and business’ behaviours affecting food waste, and by allowing the detection of intangible food waste drivers. Such an objective is achieved through the development and the testing of Agent-Based Models (ABMs) and Bayesian networks (BNs).
Matthew Grainger, Gavin Stewart, Simone Piras, Simone Righi, Marco Setti, Matteo Vittuari, 2018 “Model integration. Integrated socio-economic model on food waste”, H2020 REFRESH, Newcastle University, Newcastle-Upon-Tyne, UK.