HORIZON-CL6-2027-02-CLIMATE-02: Compound shock propagation and early warning for agri-food chains with knowledge-guided causal AI (TAGEM, Türkiye)

The AI and Agricultural Technologies Research Center of TAGEM, the research directorate of Türkiye's Ministry of Agriculture and Forestry, would like to join a CLIMATE-02 consortium in Area A (agri-food chains). We can lead a task on shock propagation modelling or on early warning tools.

The topic asks how shocks propagate along agri-food supply chains and how climate change and trade compound or amplify them. We are developing a model that addresses this question directly. A knowledge base of the supply system, covering which inputs feed which products, which routes and trade partners link markets, and the expected sign and lag of each link, is turned into a causal graph. A transformer model then estimates, only along that graph, how much a given shock changes a price compared with a situation without the shock.

The model covers geopolitical, energy, climate and weather, logistics, trade policy and macro-financial shocks. When two shocks overlap, such as a drought during a conflict, it estimates their interaction, and only where they share a channel. Where the data cannot separate the shocks, it reports a range or says that the effect is not identified, instead of giving a misleading number.

Alerts are triggered from news and data streams in near real time and use only data that were already published at that moment. Every estimate can be traced back to the chain of links it used, for example gas to fertiliser to wheat to flour, so that policy staff can check the reasoning. A language model only writes the alert text from that trace.

In the project we can apply this to EU agri-food chains such as cereals, fertiliser and feed, with Türkiye as a case between the Black Sea exporters and the EU, and test it on past episodes such as 2022. With the Ministry we can validate the alerts with policy staff.

We are looking for a coordinator, supply chain and systemic risk researchers, trade economists, social science teams and public authorities that would use early warning.