20236702139829

Overview: Agriculture and forestry provide food, feed, fiber, fuel, lumber products, and environmental services while sustaining rural and urban economies. But US global competitiveness and nutrition security are at risk due to rising greenhouse gas (GHG) concentrations, resulting in climate change, degrading ag-forest system health, and an aging and skill-deficit workforce.

To address these challenges, we propose to create a climate-focused agriculture-forestry-AI (AGFOAI) discipline, a community of practice, and functioning GHG markets by improving understanding of trade-offs and feedback loops between climate change mitigation and adaptation and between biomass productivity and GHG fluxes. We will develop AI-enhanced GHG and biomass estimation methods and spatially-explicit multiscale (field-to-market) decision support tools for equitable adaptation and mitigation.

AI advances will include reliable, accurate out-of-sample prediction [55] from sparse ground-truth measurements with consideration of hard constraints, uncertainty, and spatiotemporal variability. We propose a virtuous cycle of discovery and inquiry in foundational AI (FAI) and use-inspired research (UIR) that considers decision-making at different scales.

FAI research includes combining learning and AI reasoning, AI-aided multi-objective decision-making, and generalization theory, along with UIR areas of GHG flux estimation, land-use and cropping system change, biomass productivity, GHG markets, multiscale decision support tools, knowledge-guided machine learning (KGML), computer-vision guided perception and analysis, and AI-guided digital twins.

Intellectual Merit: Our proposed research will advance climate-smart agriculture and forestry (CSAF) knowledge and understanding to create CSAF decision support systems using KGML for reliable out-of-sample prediction [55] (AI) in un- or under-sampled fields and parcels, and AI-aided multiscale and multicriteria decision support tools for evaluating tradeoffs between alternative CSAF practices for GHG mitigation and adaptation under current and future climate scenarios. It has the potential to transform machine learning from a soft-constraint (e.g., regularizers) and mono-objective (e.g., prediction accuracy) paradigm to confront hard constraints (e.g., mass and energy balance) and multiple objectives (e.g., decision making, prediction accuracy and domain interpretability, equity, economic return, and ecosystem services). Like ImageNet [174], it has the transformative potential to advance computer vision from a human-visible spectrum and point-cloud-based approach to a sensor-rich (e.g., optical, thermal, microwave) approach by publishing new CSAF_ImageNet benchmark data and use cases (estimate GHG fluxes, soil organic carbon, biomass productivity). Our core team has a history of synergistic research and the required skills, expertise, and access to data and sensor resources.

To foster strong interactions across proposed research areas, workforce development, and collaboration nexus activities, a dedicated AI institute is needed to integrate the expertise of investigators from diverse disciplines and institutes in close collaboration with stakeholders to cultivate a new AGFOAI discipline and community of practice.

Broader Impact: The proposed institute will benefit society by catalyzing an AGFOAI discipline, a community of practice, and better functioning GHG markets. It will enhance the national research and educational infrastructure by sharing curated datasets and easy-to-use multiscale decision support tools, including AI advances (e.g., KGML, AI-guided multi-objective optimization). It will grow the American AI workforce via the integration of AGFOAI research with education; mentoring of professional, post-doctoral, graduate, and undergraduate students; engagement of secondary school teachers and students; and co-development and training of farmers and foresters in the use of AI-inspired tools, with careful consideration of broadening participation via recruitment, retention, and placement of all program participants. The team includes minority-serving institutions as active participants in all activities. Community-building activities include shared data and tools, integration of partners, and knowledge transfer via co-creation, industry consortia, and the IP framework.

Regents Of The University Of Minnesota

10.310 - Agriculture and Food Research Initiative (AFRI)

Institute of Food Production and Sustainability (IFPS) [USDA - NIFA]

Minneapolis, Minnesota 55455-2009 United States

Single Zip Code

None

Amendment Since initial award the End Date has been extended from 05/31/24 to 05/31/26 and the total obligations have increased 200% from $4,000,000 to $12,000,000.