LIRA-AI: Landscape intelligence for regeneration and adaptation

Multifunctional landscapes sustain food production, biodiversity, water resources and livelihoods, yet decisions affecting these functions are frequently made through separate sectoral processes. This fragmentation can shift environmental pressures between locations, overlook community constraints and direct investment toward interventions whose benefits may not persist under climate change. Across eastern Africa, the interdependence of farming, pastoralism, forests and upstream–downstream water users makes these coordination failures particularly consequential. Addressing them requires a shared decision process that connects scientific evidence with stakeholder priorities, implementation capacity and investment. LIRA-AI—Landscape Intelligence for Regeneration and Adaptation—is an agentic decision-support prototype developed to support this connection. Its strategic contribution is to organize complementary scientific assessments around a common landscape problem, using degradation diagnosis and regeneration planning as the initial entry point. Environmental evidence informs diagnostic classifications, which generate candidate intervention packages. Climate-scenario information then provides context for examining how changing rainfall, dry spells, heat and precipitation extremes could affect the suitability and durability of proposed measures. Community, policy and tradeoff assessments bring additional perspectives on feasibility, competing objectives and the distribution of co-benefits and burdens into human review. This proof-of-concept makes the connection between understanding landscape conditions and deciding what action is appropriate, where, and for whom more explicit. Investment passports carry the resulting rationale, assumptions, indicative costs and monitoring requirements into project preparation, while communication and monitoring functions provide a basis for implementation support and subsequent review. This architecture is intended to reduce fragmentation across the decision cycle; effective coordination and improved outcomes remain to be demonstrated. A computational demonstration using four stored records associated with highland, midland, pastoral and riverine settings in Ethiopia’s Omo-Ghibe MFL landscape reproduced selected diagnostic and intervention-generation routines. Sensitivity to assumed inputs and missing-data treatment identified priorities for validation, rather than establishing diagnostic accuracy or field effectiveness. LIRA-AI pilot is a foundation for connecting CGIAR Multifunctional Landscapes priorities for integrated management and inclusive planning with Climate Action priorities for climate-risk management and locally led adaptation. Sharing its source code enables scientific scrutiny and collaborative extension. Further investment would test whether this coordinated approach improves decision quality, planning efficiency, intervention feasibility and institutional use, establishing the basis for evaluating more durable and equitably distributed landscape benefits.