Within the EcoWheataly project, there is a crucial technological core: an agent-based simulation model (ABM) capable of representing thousands of Italian wheat farms — both real and virtual — to study how agricultural policies may shape the future of durum wheat production in Italy.
The starting point is simple to describe but complex to build: reconstructing an agricultural Italy that faithfully reflects real data, combining multiple sources (RICA and the ISTAT Census) and filling the gaps using advanced statistical algorithms.
From data to simulation: the load balancing that builds the farms
The process begins with two datasets:
- RICA, which provides data on real farms
- ISTAT, which provides the number of plots for each province × altimetric zone
These two worlds do not perfectly coincide: in some areas there are more plots than observed farms, in others the opposite.
For this reason, EcoWheataly applies a load balancing algorithm: a procedure that compares RICA and ISTAT territory by territory and calculates how many farms are “missing” to achieve a complete representation.
The result?
- 1846 real farms,
- 242 artificial farms,
- for a total of 2088 simulated farms, distributed across 4 parallel computational ranks, each containing both real and artificial farms.
The number of artificial farms (242) is kept to a minimum (2 per province-elevation) for testing purposes.
At full scale, there will be about 130,000 farms; that is, the farms active in durum wheat production according to the census.
This distributed structure makes it possible to handle large simulation workloads while maintaining territorial and statistical coherence.
How artificial farms are created
To generate farms that do not exist in the RICA dataset, the team uses a configuration file built from statistical estimates.
For each province and altimetric zone, it includes:
- average cultivated areas,
- farmer characteristics,
- farm size,
- agronomic variables,
all reconstructed from the distributions observed in the official datasets.
When a value is missing, it is drawn from a purpose-built statistical distribution — an elegant way to obtain farms that are realistic and consistent with the territory.
Aligning RICA to the model: the case of farmer age
A key example concerns age: RICA provides only a binary classification (“young” ≤ 40, “not young” > 40), while the decision model requires a numerical age.
For this reason, age is reconstructed using a statistically estimated distribution based on available data — an essential step to make each agent’s behaviour more realistic.
Real agents and virtual agents: a complete agricultural Italy
Once load balancing is completed, each simulation rank contains:
- a file with real farms (with all observed variables)
- a file with artificial farms (with variables drawn from RICA/ISTAT distributions)
Together, these two groups replicate a complete, statistically robust map of wheat farming in Italy.
In the model, there is no difference between real and virtual agents: both follow decision rules, respond to policies, and compete for resources and yields.
Inside the decision-making process
The heart of the project is the farmers’ decision-making process, which integrates:
- socioeconomic characteristics,
- agronomic variables,
- territorial constraints,
- responses to public policies
Each agent makes decisions — from fertilizer use to land management — based on a model inspired by microeconomics and sustainability, already developed in the group’s broader research.
Possible developments of the model
Future developments of the ABM system include:
- integration of additional production inputs,
- new environmental impact indicators,
- greater territorial detail,
- calibration with updated RICA and ISTAT data,
- expansion towards more complex policy-response models.
Conclusion: a simulation that gives voice to Italian agriculture
The EcoWheataly ABM is not just a technical exercise; it is a way to let thousands of Italian farms “speak” — including those missing from the datasets but necessary for representing the territory.
Thanks to this faithful reconstruction, the project can simulate CAP scenarios, evaluate sustainability policies, and understand how Italian farmers may genuinely respond to incentives.
A fundamental step toward designing policies that truly work, on real fields.
Sources:
- Giulioni, G. (2025). Integration of RICA data, ISTAT data, and the decision-making model – EcoWheataly. Presentation at the EcoWheataly Meeting, Rome, 11 November 2025. Department of Socio-Economic, Managerial and Statistical Studies (DiSEGS), University “G. d’Annunzio” of Chieti-Pescara.