CLIM: Transition Workshop - Agent-Based Modeling for Food Systems - Adway Mitra, May 14, 2018
1. Agent-Based Modeling
for Food Systems
Adway Mitra Food Systems Working Group, SAMSI
(with Hans Kaper, Hans Engler, Kaitlin Hill, Mary Lou Zeeman, Margaret
Johnson, Huang Huang, Jim Rosenberger)
2. Agenda of the Talk
Introduction to Agent-based Models
Agent-based Models for Food systems
Future direction: Agent-based Models + Food Networks+ Climate Networks?
3. Modelling Complex System Dynamics
A system of multiple “agents” who interact among themselves and with the
“environment”
Eg. a traffic crossing with cars, pedestrians and signals
Complexity of system: number of agents, types of agents, interactions
Equation-based Models:
System described at each time-point using state-variable
Eg. number of cars on a traffic crossing at any moment, signal colors
System evolution described using state variables and state transition equations
Eg. how the number of cars change depending on the signal
Amenable to detailed mathematical analysis (eg. steady-state
computation) but can model very few interactions, cannot model agent-agent
interactions)
4. Agent-based Modelling
Models the behaviour of individual agents instead of the entire system
An agent is capable of controlling its own decision-making and acting, based
on its perception of its environment, in pursuit of one or more objectives
Model Dynamics: agent behaviour, agent-agent interaction, agent-
environment interaction
Allows us to visualize and understand highly complex systems through
simulation
Can simulate highly complex systems, but not amenable to rigorous
mathematical analysis
5. Aims of Agent-based Modelling
Emergent Behaviour: How complex behaviour evolves
or emerges from relatively simple local interactions
between system components over time ?
- Are there frequent traffic jams at the crossing?
Spontaneous traffic gridlock emerges
6. Aims of Agent-based Modelling
Policy impact assessment: Can external interventions on
the environment or on the interaction rules affect the
system’s evolving patterns?
- Does changing the signalling scheme, crackdown on
signal violators improve the situation?
Longer phases at traffic
lights lead to worse gridlock
7. Agent-based Modelling in a Food System
Miller et al, Using stylized agent-based models for population-environment research: a case study from the Galápagos Islands,
Population and Environment, 2010.
Scenario: Predatory grape plants covering farmlands in Galapagos Island,
forcing farmers to either clean up fields or change occupation
Agents: Farmers with variable property and skill levels
Environment: Farmlands and invasive grape plants (can be modelled with EBM)
Agent actions: make choice based on individual’s attributes (clean up or switch profession)
Emergent behaviour: reduction in farm output
Policy interventions: Provide limited financial assistance to clean up fields
8. Agent-based Modelling in a Food System
Scenario: A Beef Production system
Agents: livestock animals, farmers, consumers, other stakeholders
Agent attributes: age/weight/health etc for livestock, wealth and skills for farmers,
health/wealth/diet for consumers etc
Interactions: between animals (herd), farmer-animal, farmer-consumer etc
Emergent Behaviour: effects on environment – reduction of green cover, nitrogen/carbon emission,
consumers’ health etc
Policy interventions: incentivizing grass-fed rearing, restrictions on transportation, dietary restriction
on consumers etc
9. Beef Production system (J.L.Capper, Animals, 2012)
This system can be modelled
using equation-based approach
Number of cattle at each stage
can be modelled using population
growth models (eg. Leslie matrix)
J.L.Capper: CON/NAT cause less harm to environment than GFD!
10. ABM for Beef Production systems
(C.W. Ross, R.J. Glass, J. Harger, et al, Development of an Agent-
Based Epidemiological Model of Beef Cattle)
Agent-based models can also
be used for this system.
Each livestock animal considered as
an agent
Suitable to model each animal’s
life, health, interactions with others
11. ABM for Beef Production Systems:
future work
ABM Modelling Challenges:
Deployment of agents for farmers, consumers and other stakeholders?
Transportation networks? Modelling of infectious disease spreading?
Policy interventions through ABM:
How can we incentivize customers to shift from beef-based diet to more
sustainable diet?
How can we incentivize farmers to shift from producing beef to more
sustainable food?
12. Modelling larger/global food systems
Macro-scale: Global food production network, using
annual food production data
Micro-scale: Agent-based model for producers, food
chain actors, consumers, etc.
Climate Network: well-studied
Can we explore the connections between global climate network
(eg. precipitation network) and global food network?
Donges et al, The backbone of
the climate network, 2009
Poladian et al, Complex Global Network Makes
Food Poisoning Outbreaks Hard to Track, 2012
13. Climate Change and Food Systems: Can
ABM help?
Climate Change as external forcing factor on Food System
Modelling challenges for ABMs:
Model the impact of climate on crops and livestock
Model the impact on consumers
Model the impact of mitigation policies on consumers, farmers and other stakeholders
Emergent Behaviour and Policy interventions:
How are food systems affected by climate extreme events?
How are the inter-agent interactions of a food system affected by climate
extremes?