Brightcon 2025, hackathon & courses in Grenoble and online

This abstract is an essential description of the work going on within the "Evaluation of policies for enhancing sustainable wheat production in Italy"
(ECOWHEATALY) research project. The ECOWHEATALY project focuses on the power of policies in providing economic incentives for farms switching from less to more sustainable wheat production techniques.
To address real-world complexity, ECOWHEATALY developed an agent-based model for the wheat production system in Italy. This model is interfaced with an existing tool that can handle international wheat markets, as well as a module to measure the sustainability of the system.
The three key components of the model framework are: a) the ABM, fed by wheat farm data and green policy actions; b) the Global Economic Model (GEM), which outputs wheat prices at major international markets starting from produced and demanded quantity at the global level; and c) the LCA module, which measures the environmental impact deriving from individual farm-level decisions. The software to run the three modules is developed using the Repast Suite (https://repast.github.io) for the ABM and the GEM, and Brightway (https://docs.brightway.dev/en/latest/) for the LCA.
Regarding the LCA, we encountered several issues during project development, particularly because the project aims to produce and utilize open resources as much as possible.
Among the software tools for performing LCA analysis, we have chosen Brightway due to its open-source nature and the possibility of using it programmatically.
One of the project's objectives is to raise awareness among agents about the sustainability of their production choices, enabling them to adopt more sustainable practices. To this aim, we implemented the LCA for a very simple wheat production process. This is mainly to keep computation lightweight because our agent-based model has about 200000 farms (which is the number of wheat producers in Italy, according to the 2020 census). In our model, farmers use the following four inputs to produce wheat: 1) Tractor power, 2) Nitrogen as fertilizer, 3) 2,4-D salt as herbicide active principle, and 4) Deltamethrin as insecticide active principle.
The inventory phase is based on open resources. The tractor power activity was sourced from the "University of Washington Design for Environment Laboratory/Field Crop Production” database, available at the USDA’s LCA Commons. We load the process's XML file with the SingleOutputEcospold1Importer function.
The Nitrogen activity was digitized from the appendix of a scientific paper (Brentrup et al. 2004) and read into Python as a csv file. Due to the lack of an open inventory analysis for 2,4-D salt and Deltamethrin, we add them directly as outputs to the wheat production process.
Unfortunately, we have no access to Ecoinvent database. Therefore, we use the biosphere3 database provided with the Brightway2 version. The matching between the tractor power and nitrogen processes and the biosphere3 database was done by fuzzy string matching.
Concerning the LCIA phase, we chose the ReCiPe 2016 methodology. After installing the bw-recipe-2016, our main effort was to update the characterization factors of the original methods with those provided for Italy, where possible, i.e., Terrestrial Acidification, Particulate Matter Formation, Ozone Formation, and Freshwater Eutrophication.
We wrote several Python scripts to perform the operations described above. They will be made available on GitHub shortly. In any case, we will be happy to share them with interested people.
The main problem we encountered when integrating Brightway2 with our Repast4py agent-based model was the failure registered during the parallel execution of simulations. The failure seems to be caused by Brightway’s need to read data from databases.
When a considerable number of agents (farmers) perform LCA simultaneously, the database is overloaded with queries. Some queries find the database busy (locked), and the simulation stalls.
Our solution was to equip agents with the necessary matrices to perform the calculation, allow them to customize these matrices, and perform the computation without querying the database. We used Brightway2 to write the Activity, Biosphere, and Characterization matrices to a file that is imported by every farmer of our agent-based model to perform the relevant matrix calculations.
We would be delighted to present our work at Brightcom 2025 and discuss potential improvements during the conference.
We also took note of our experience in learning Brightway in a 22-page document titled "A basic introduction to LCA with open resources". Currently, it is a LaTeX document, but we are willing to provide it as an IPython file if it will be judged helpful to Brightway beginners.