1 of 35

AI-CLIMATE: A National AI Research Institute (NAIRI) on

Climate-Land Interactions for Mitigation, Adaptation, Tradeoffs and Economy

Director: Shashi Shekhar

Lead Institution: University of Minnesota

�

Acknowledgements: NSF , , USDA/NIFA

The AI-CLIMATE is one of 7 National AI Research Institutes (NAIRIs) announced on May 4^th, 2023. These Institutes aim to catalyze collaborative efforts across institutions of higher education, federal agencies, industry, and others to pursue transformative AI advances that are ethical, trustworthy, responsible, and serve the public good. Also, they bolster AI R&D infrastructure and support the development of a diverse AI workforce. They will drive breakthroughs in critical areas, including climate, agriculture, energy, etc.

AI-CLIMATE

2 of 35

Agriculture Today: Societal Importance

Agriculture nourishes us with

Food, Feed, Fiber, Fuel

Economic Opportunities

10% of U.S. Jobs
Early adopter of technology, e.g., GPS, UAV, …

Steward of natural resources

Conservation, Water quality, Soil health, …

Curb climate change

Traditional: Reduce soil emissions (25% of MN and 10% of US emissions) + Biofuels …
Carbon removal: Soil, forest, nature can remove ~ 21% of US (33% global) Emissions

AI-CLIMATE

3 of 35

Agriculture Today: Challenges

Environmental

Climate Change
Soil degradation, Bee colony collapse

Cyber

Sensors, Data, Algorithms, Models, Tools, Automation
Crop yield/stress, soil health, …

Social

Population growth
Aging workforce, Labor shortage

AI-CLIMATE

4 of 35

Situation and AI-driven Natural Climate Solutions

By 2050, the United States aims to have net zero carbon emissions, and one of the most promising ways to do this is using natural systems like forestry and agriculture as ‘carbon sinks.’

Mitigation Potential

21% of current net annual U.S. emissions [1]
A third of global emissions [2,3]

Natural Climate Solutions (a.k.a. Climate Smart Practices)

Forest: reforestation, forest & fire management
Agriculture: cover crops, nutrient management, biochar, alley cropping
Wetlands, Grasslands,

Key Decision

Which natural solution to use where? when? how long?

Role of AI = Inform the Key Decision

Better Data, Models, Algorithms, Tools

Sources:

[1] Natural climate solutions for the United States, Science Advances, 4(11), Nov. 14th, 2018.

[2] Natural Climate Solutions, nature.org (TNC)

[3] C. Girardin, et al., Nature-based solutions can help cool the planet, nature, 12 May 2021.

AI-CLIMATE

5 of 35

Vision

Transform the science of AI and decision support tools for climate-smart practices in agriculture and forestry to co-create solutions for previously unsolvable problems, and accelerate adaptation to and mitigation of climate change, while informing policy and empowering carbon markets.

This transformation is built on:

WHAT: Next-generation AI-CLIMATE and AI-guided decision support tools
HOW: User-guided co-creation
WHO: Unprecedented integration

AI-CLIMATE

6 of 35

Participating Institutions

Center for Agricultural Resources Research (CARR)

Government

Institutional Support & Legacy

Example Industries, NGOs

AGRONOMY E-LEARNING ACADEMY

Member Institutions

AI-CLIMATE

7 of 35

Team Members

Artificial Intelligence (AI)

Broader Impacts

Broadening Participation, Education & Workforce Development, Collaboration & Knowledge Transfer

Cyber Infrastructure

Applied Economics

Climate Smart Agriculture and/or Forestry (CSAF)

AI-CLIMATE

8 of 35

Institute Approach

MISSION Establish an AI-CLIMATE discipline, innovation ecosystem & community of practice

VISION Thriving AI-powered climate-smart practices & greenhouse gas markets

Stakeholder Needs

Data, models, tools
Workforce
Boost GHG markets
Inform policy

Management, Integration

Research

Education & Workforce

Collaboration & Knowledge Sharing

Channels

College

Professionals

K6-12

Co-creation

Deployment

Adoption

Democratization

Data sharing

Software Tools

Cyber-�infrastructure

Institutional Values and Value-Propositions

Mission-Focus

Team Science

Project Selection

Anticipate Risks

Broadening Participation and Stakeholder Engagement

AI-CLIMATE

9 of 35

Needs Assessment

Carbon markets

More accurate Soil-Carbon estimation

2022 Survey of Land Stewards

Know climate-smart practices (CSPs)
Q? Which CSP to use where? when?
Ex. 7-mile creek watershed, MN

Conservation tillage with stover removal

Low phosphorous application

Prairie grass

Switchgrass

Conventional tillage

Conservation tillage

7-mile Creek Watershed, MN

23,552 acres, 36.8 sq. miles

Sediment: 2585 ton/yr

Unchangeable landscape

Public water

Watershed outlet

Watershed boundary

Details: Y. Xie, B. Runck, S. Shekhar, L. Kne, D. Mulla, N. Jordan, and P. Wringa, Collaborative Geodesign and Spatial Optimization for Fragment-Free Land Allocation, ISPRS Int. J. Geo-Inf. 2017, 6(7), 226; https://doi.org/10.3390/ijgi6070226.

AI-CLIMATE

10 of 35

Research Needs

AI Research Themes (T)

Integrative Vignettes (V)

Better GHG, SOC, CDR estimation & verification
Climate risks, adaptation, & shifts in cropland-forest transitions
Multi-criteria optimization of mitigation practices & productivity.
AI-guided emulation of earth-economy macroeconomic ecosystem service payment markets.
Multi-scale multi-criteria GHG decision support tools.

Out-of-sample Prediction (variability, sparse ground truth)
Hard constraints, mechanistic models (e.g., mass or energy balance)
RGB+Lidar (hard-to-handle rich spectral imagery, data variety)
Multiple objectives (e.g., economic, eco-services, equity)
Trade-offs, feedback loops

Knowledge Gaps (KG)

T1. UIR: Knowledge-Guided Machine Learning (KGML) for GHG & C-Cycle modeling
T2. FAI: Combining Learning & AI Reasoning (CLeAR)
T3. UIR: Computer Vision Guided Perception and Analysis (CVPA)
T4. FAI: AI-aided Multi-objective Optimization for CSAF Decision-making (AIMOD)
T5. UIR: AI-aided Digital Twins (AIDT) to facilitate resilience planning for climate scenarios

AI-CLIMATE

11 of 35

Innovation Aims

Better Data, e.g., finer resolution maps of soil- moisture, organic carbon
Multi-scale Multi-criteria Decision Support Tools for GHG mitigation
More accurate Models of soil GHG emissions
Scalable Algorithms, e.g., multi-objective optimization

AI-CLIMATE

12 of 35

AI for Better Data, e.g., Sharper Maps

Need: Soil moisture map
Challenge: Satellite map too blurry
?under-water dry-spots, flood wet-spots
AI can make sharper maps
Map Soil moisture (at 5cm depth of top soil near roots)

NASA SMAP Satellite (9 Km Resolution, July 16-17 )

AI + Ground Sensor (30 m Resolution, July 16)

Details: P. Khandelwal, et al., DeepSoil: A Science-guided Framework for Generating High Precision Soil Moisture Maps by Reconciling Measurement Profiles Across In-situ and Remote Sensing Data, ACM SIGSPATIAL 2024.

AI-CLIMATE

13 of 35

B. Multi-scale Multi-criteria Decision Support Tools

Tools	Stakeholder	Scale	Example Decisions
AI-COMET Farm	Land Stewards	Farm, Forest Stand	Compare climate smart management practices
AI-GeoDesign	Groups	Watershed	Social learning of spatial interactions and tradeoffs
AI-Earth-Economy, Soils-Revealed	Policy Makers	Country, State, Watershed	Compare policy interventions and ecosystem service tradeoffs
All	Companies	All	Cost and payment for each practice

AI-CLIMATE

14 of 35

C. AI for Better Prediction Models: Greenhouse Gas Emission

Need: Predict land emissions (to curb climate change)

Challenges: Out of sample prediction, spatial variability, …

Approach: Knowledge Guided Machine Learning (KGML) = AI + Laws of Nature + Sensor Data

Impact: Improved prediction accuracy

Sources: Prof. Z. Jin, and Prof. V. Kumar, University of Minnesota

Details: Knowledge-guided machine learning can improve carbon cycle quantification in agroecosystems. Nat Commun 15, 357 (2024).

AI-CLIMATE

15 of 35

Institution Approach

Team

C. More Accurate Models: Knowledge-Guided Machine Learning

Use Case: Accurate quantification of soil GHG emission

Knowledge Gaps: Out-of-sample prediction, data paucity, …

Approaches

Model complex physical process
Knowledge-guided transfer learning
Assimilate remote sensing data
Represent Crop x Environ x Mgmt impacts on GHG

Basis of confidence

Modeling of physical processes in hydrology
Improved modeling accuracy for GHG
Scales up GHG models to broad regions

KGML is capable of representing multiple interacting biophysical processes and advances frontier

Mass Balance

Details: Knowledge-guided machine learning can improve carbon cycle quantification in agroecosystems. Nat Commun 15, 357 (2024).

AI-CLIMATE

16 of 35

Institution Approach

D. Scalable Algorithms: AI-aided multi-objective decision-making

Use Cases

Multi-criteria (economic, ecoservices, equity, resilience) decisions
Tradeoff analyses and decision making

Knowledge Gaps

Hard and combinatorial constraints
Diverse solutions wrt tradeoffs (beyond single objective optimization.

AIMOD approaches

Exact and approximation algorithms for Multi-objective Pareto optimization (e.g., DP based alg. with O(n log n) pruning)
Enforce hard constraints (e.g., mass balance)

Outputs

M. Grimson et al., Scaling Up Pareto Optimization for Tree Structures with Affine Transformations: Evaluating Hybrid Floating Solar-Hydropower Systems in the Amazon (Best Student Paper for the AI for Social Impact track at AAAI-2024)
Amazon basin (5M+ river segments, 6 objectives) - Pareto frontier (approx. algo. with guarantees) for hydropower, GHG emissions and eco-services tradeoffs (Science 2022)

AI-CLIMATE

17 of 35

Institution Approach

D. Scalable Algorithms: Ex. 7-mile Creek Watershed

Manual Collaborative Geodesign

Multi-objective Optimization Algorithms

AI-CLIMATE

18 of 35

Collaboration and Knowledge Transfer

AI-CLIMATE will Build Partnerships with a Diverse Range of External Stakeholders

Create

a high-performing, diverse and passionate team with a shared vision

Leverage

and build on existing resources, e.g., datasets, software tools, CI, AI artifacts

Evolve

to provide value to broad set of stakeholders and respond to emerging opportunities

AI-CLIMATE Network of Networks

Educators

Government Agencies

Research Lifecycle

Farmers

Foresters

AI-CLIMATE Businesses

NGOs

Future Workforce

USDA Climate Hubs

Climate Smart

Commodities

AI-CLIMATE

19 of 35

Collaboration Nexus Opportunities

Use-case: Improve Soil carbon and emission estimates
Situation; Better data with industry but better models with academia
Opportunity: Synergize and disseminate better data and models

Industry (FFAR) data & users

More accurate Models of soil carbon & emissions

User community, e.g., Climate Smart Commodities

AI-CLIMATE

20 of 35

Collaboration: Co-creation Activities

Team: Climate-Smart Commodity Projects, Companies, Non-profits, Policy makers, Stewards
Co-Visioning: surveys, focus group, meetings

Curb climate change by enabling land stewards, markets and policymakers
Challenges: Gaps in data, knowledge, science, tools, …

Co-select Questions: Which climate smart practice should be used where and when?
Co-Discoveries: Cheaper & more accurate estimation of carbon sequestration in soil, trees, …
Co-Inventions: Decision Support Tools for land stewards and policy makers
Co-Evaluation: COMET Farm users, …
Edu. & Workforce: FFA, 4H, eLearning Academy, Digital Ag minor/major, ..
Impacts (expected): Inform Policy (MN, US), Accelerate Adoption, Lubricate Carbon market, …

AI-CLIMATE

21 of 35

Value Propositions of AI-CLIMATE

System of systems approach

Cross-cutting many choices

Agile co-creation

Partners with broad reach

Climate change is an urgent and critical global challenge and addressing it will require a portfolio of responses

Bigger potential impact

System of systems approach
Solve previously unsolvable problems leveraging and advancing AI and climate smart ag and forestry
Considering multiple dimensions (e.g., goals, management practices, Integrated Research Education Extension)

Agile co-creation

Influential partners for deployment, adoption, and democratization
Leading scientists with long record of AI collaboration in agriculture and forestry
Established collaborations across institutions in research, education, and Extension

AI-CLIMATE

22 of 35

AI-CLIMATE Activities

AI Leadership Summit

Internal

Retreat, strategic implementation plan, annual review
Website https://cse.umn.edu/aiclimate

External Visibility

Hill AI Institute Day
Visits by NSF Director and USDA Secy.
Summit on AI Leadership (AI Institute PI Workshop)

USDA Secy. Visit

May 2024 Annual Meeting

NSF Director Visit

Hill AI Institutes Day

AI-CLIMATE

23 of 35

Outreach

Policy Makers

Federal

Need: Further engage NIFA/USDA leaders with AI
Senate and House Agriculture Committee Staffers, Sept. 20th, 2023.
Booth at Hill AI Institute Day, Senate Office Building, Sept. 19th, 2023
Senate Judiciary Committee Chief Counsel (Avery Gardiner) and Tech. Fellow (Divya Goel), June 1st, 2023
Census Bureau Geography Division Chief Deirdre Bishop, Dec. 5th, 2023

State (Minnesota House of Representatives)

Rep. Larry Kraft (Dem.), Rep. Bjorn Olson (Asst. Minority Leader, GOP), Rep. Kristi Pursell (Dem.)
Multiple meetings on soil carbon sequestration

Others

US:NSF - India:TIH PI Meeting, Baltimore, May 22nd-23rd, 2023
Carbon Science Talks, Center for Carbon Research in Tropical Agriculture, Univ. de Sao Paolo, Brazil (Oct. 11th, 2023)
Bezos Earth Fund, AI for Climate and Nature Workshop (Oct. 17th-18th, 2023)
Environmental Defense Fund to co-organize a Workshop with AIFARMS
MN Farmers Union : Midwest Climate Smart Commodities Grantees Fall 2023
Sierra Club: North Star Chapter, Forestry Group
sister NIFA sponsored AI Institutes: Planning a Workshop on AI for Ag. (July 2024),
Companies: MBOLD (2/22/2024), TrueTerra, Exxon Mobil, …

AI-CLIMATE

24 of 35

Soil Emissions: Greenhouse, Biochar Perennial Crops Soil Emissions: Field Measurement

Kernza

Camelina

N2O peaks after big rains

Flux Chamber

Soil Greenhouse

Biochar

25 of 35

Summary

Agriculture and Forestry are important but facing challenge

Land Stewards know Climate Smart practices

Q? Which climate smart practice should be used where? When? How long?

How can AI help?

Improve Data, e.g., finer resolution maps of soil moisture, organic carbon, …
Better Decision Support Tools
More accurate Prediction Models, e.g. for soil GHG emissions
Scalable Algorithms, e.g., multi-objective optimization

How may AI benefit? : Strengthen AI for Science

Knowledge-Guided Machine Learning : honor physical laws

AI-CLIMATE

26 of 35

National AI research Institute for Climate-Land Interactions, Mitigation, Adaptation, Trade-Offs, and Economy, NIFA 2023-03616, $20M, 6/2023-5/2028 (P.I.: Shashi Shekhar). cse.umn.edu/aiclimate

AI-CLIMATE

The AI-CLIMATE is one of 7 National AI Research Institutes (NAIRIs) announced on May 4^th, 2023 under the federal responsible AI initiative. These Institutes aim to catalyze collaborative efforts across institutions of higher education, federal agencies, industry, and others to pursue transformative AI advances that are ethical, trustworthy, responsible, and serve the public good. Also, they bolster AI R&D infrastructure and support the development of a diverse AI workforce. They will drive breakthroughs in critical areas, including climate, agriculture, energy, etc.

NSF Director Visit

May 2024 Annual Meeting

AI-CLIMATE

27 of 35

Innovation Goals

More accurate Models, e.g., soil emissions
Scalable Algorithms, e.g., multi-objective optimization
Better Data, e.g., finer resolution maps of soil- moisture, carbon
Multi-scale Multi-criteria Decision Support Tools for GHG mitigation

Potential Impacts

Strengthen AI for Science (e.g., honor physical laws)
Mitigation: Accelerate Carbon-sequestration in farms and forests
Adaptation: Drought resilience via healthier soil
Economy: Empower Carbon markets by better carbon-accounting
Expand and diversity AI-ready CSAF workforce

News

AI could help track farm country's carbon emissions, U study says, Star Tribune, March 21st, 2024.
newsclip video (2 minutes) in U of M looks to use AI to create more climate-safe farming practices, KARE11.com, May 16th, 2023.

AI-CLIMATE: Innovation Goals

GeoDesign Tool

Short Video

AI-CLIMATE

28 of 35

Pictures

Blog

AI-CLIMATE

29 of 35

AI-CLIMATE External Activities

May: interview with KARE11 ( news, video )
June 1st: Meet with Senate Judiciary Committee Chief Counsel
Sept. 7th: Meet with MN Legislators Larry Craft, Bjorn Olson, Kristi Pursell
Sept. 18th-20th: Hill AI Institute Day (video to share with visitors)
Oct. 13th: Visit by NSF Director
Oct. 16th: UMN CSE 9-9-9 Workshop on Ag/Bio, Meeting with Gevo
Oct. 17th-18th: Bezos’ Earth Fund AI for Climate Workshop (presentations)
Oct. 23rd-27th: Summit on AI Leadership (AI Institute PI Workshop)
Oct. 31st: Presentation to Jack Dangermond, President ESRI
Nov. 13th: Meet with NASEM Climate Crossroad (Amanda Staudt, Alex Reich)
Nov. 18th: Panel Presentation and Poster at Computer Science Research Showcase
Dec. 5th: Presentation to Deirdre D. Bishop, Census Bureau Geography Division Chief
Dec. 2023: UMN RIO: Sustainable Futures: … Geodesign … for Data-Driven SDG … Action
Jan. 2024 UMN Provost with SDG Group
Feb. 2024 MBOLD companies (via UMN RIO)

AI-CLIMATE

30 of 35

Hill AI Institute Day

Sept. 18th, 2023 : NSF AI Institute Day, NSF HQ
Sept. 19th, 2023 : Hill AI Institute Day, Senate Office Building

NSF Director, NSF CISE AD and PMs, …, Staff of NIFA Director, OMB, etc.; Lewis Burke Assoc., CNSF, …
Natl. Endowment Humanities Research Center for AI RFP (Meaghan Brown), Am. Math. Society (Karen Saxe), …

Sept. 20th, 2023: Meetings with Senate and House Ag. Committee Staff

AI-CLIMATE Booth

(Prof. Heidi Roop and

Prof. Shashi Shekhar)

Soil Samples

(Prof. David Mulla)

AI-CLIMATE Up and Running - Shashi Shekhar

AI-CLIMATE

31 of 35

NSF Director : UMN Visit

Oct. 13th, 2023

Morning: NSF Director highlighted AI-CLIMATE in his public presentation at the town hall
Afternoon: AI-CLIMATE made a 20-minute presentation to the NSF Director

AI-CLIMATE Up and Running - Shashi Shekhar

AI-CLIMATE

32 of 35

Summit on AI Leadership (PI Meeting)

Oct. 25th, 2023: Panel on AI for Scientific Discovery (Prof. Vipin Kumar)
Oct. 26th, 2023: Booth (Prof. Raju Vatsavai and Prof. Shashi Shekhar)
Oct. 26th, 2023: Panel on Ag., Digital Ag. and Health (Prof. Shashi Shekhar)
Meetings with program managers Steven Thomson (NIFA) and Jim Donlon (NSF)

AI-CLIMATE Up and Running - Shashi Shekhar

AI-CLIMATE

33 of 35

AI Opportunities

National Science Foundation AI webpage, AI Funding Opportunities Search Result

23-610: National AI Research Institutes: Astronomy (Feb. 12th, 2024), Materials Research (May 17th, 2024), …
24-518: Collaborations in AI and Geosciences (CAIG), Mar. 15th, 2024 (webinar: M, Jan. 22, at 12noon)
24-039: DCL: Funding Opportunities for Engineering Research in AI, April 30th, 2024.
23-614: Smart Health and Biomedical Research in the Era of AI and Adv. Data Science (SCH), Oct. 3rd, 2024.
24-045: DCL: Funding Opportunities for Engineering Research to Achieve Net-Zero Climate Goals by 2050
23-606: Math. Sc. Research Institutes, March 14th, 2024; 23-594: Geoinformatics , Dec. 6th, 2024 (recurring annually).
21-536: NSF Research Traineeship Program (NRT), Sept. 6th, 2024.
24-025: DCL: Advancing education for the future AI workforce (EducateAI) for UG and K-12

USDOD

NGA Research BIG-ST BAA: Topic # 01 – Geospatial-Intelligence Foundation Model (GFM), Mar. 4th, 2024
DARPA-PA-23-04-01: AI Exploration (AIE): Foundation Models for Scientific Discovery (FoundSci), Dec. 2023.

Other

National Endowment for Humanities : Humanities Research Centers on Artificial Intelligence, Feb. 14th, 2024.
State of Minnesota: Legislative-Citizen Commission on Minnesota Resources (LCCMR), usually due late March
UMN Data Science Initiative Small Seed Grant, Feb. 15th, 2024.

AI-CLIMATE

34 of 35

CSAF Opportunities

Big levers to accelerate climate action to “speed and scale”

Innovations/Tech., Carbon markets, Policy, e.g., GHG Footprint reporting, Investments

Govt. Investments

USDOE etc. Clean energy (e.g., biofuel): $100B+
USDA/NIFA climate smart commodities: $3.1B with 60M+ grants
NSF Regional Innovation Engines ($160M grants)

Philanthrophic Investments

Bezos Earth Fund: $10B (1.84B granted, 100M-400M grants)

Conserve & Restore Nature; Future of Food; Environmental Justice
Decarbonize Energy & Industry ; Economics, Finance & Markets,
Next Technologies; Monitoring Data and Accountability
Workshop on “AI for Climate and Nature”, Oct. 17th-18th, 2023.

Waverly Street Foundation (Laurene P. Jobs) : $3.5B for high-risk climate projects

AI-CLIMATE

35 of 35

Opportunities

USDA Science and Research Strategy (2023-2026), May 2023.

Accelerate Innovative Technologies and Practices (e.g., Technology/AI-enabled collaborative Decision Support Systems)
Driving Climate Smart Solutions (e.g., Impacts, Mitigation, Adaptation, Bioeconomy)
Bolstering Nutrition Security and Health (e.g., Predictive Analytics, Pathogen Virulence Factors, …)
Cultivating Resilient Ecosystems (e.g., Genomics, Genome Editing, Microbiome, Biodiversity, Infectious Diseases/Pests)
Translating Research into Practice (e.g., Education and Workforce Development, …)

NIFA : Program overview 2021 slides

Ex, A1511: Nano-tech., A1521: Eng. for Ag. Prod./Proc., A1531: Bio-Refining/Manf., Data Sc. for Food & Ag., (DSFAS), Sus. Ag. Res. & Ed. (SARE), Robotics (NRI), CPS, AI Inst. (NAIRI) Ag Genome to Phenome, …,

NSF (with USDA NIFA)

NSF DCL 23-027: Supporting Research on Food and Ag Challenges Facing Local Communities through the Civic Innovation Challenge
National AI Research Institutes (AIFARMS, AIFS, AgAID, AIIRA, AI-CLIMATE)
NSF 22-550: Signals in the Soil , Plant Biotic Interactions (PBI)
NSF DCL 19-051: Supporting Research at the Intersection of Agricultural Science, Big Data, Informatics, and Smart Communities

Cyber Physical Systems, CISE/IIS Core Programs, Smart and Connected Communities

USDOE

Bioenergy Technologies Office Funding Opportunities
EERE, ARPA-E, …

AI-CLIMATE