Nextgen Agricultural System Models and Use Cases: the Economic Data Challenge

Summary:

• AgMIP: Global Agricultural Model Intercomparison project

• Bridges gap between data, models and users of agricultural models

• Inspired by the Climate Model Intercomparison Project (CMIP)

• Areas:

  • Sustainable Farming

  • Crop model intercomparison

  • Global and Regional Agriculture Assessments

  • NextGen Knowledge Products, Models and Data

• Flow

  • Use-cases (small farms, commercial farms, policy)

  • Pre-competitive data and models (shared by community)

    • Base models, understanding of crops and pests

  • Competitive data and knowledge products

    • Early warning, ag monitoring, nutrition tracker, mobile sensors and big data

• Hard problems: agricultural practices, crops very heterogeneous

  • Yields in US vary a lot, globally much more

• Agricultural systems are very complex, design requires many tradeoffs

• While improving sustainability, must meet farmers’ profit goals (they can’t go bankrupt in the process)

• Economic analysis of agricultural systems

  • Internal validity: captures a single economic systems

  • Prediction in a different but observable environment

  • Prediction in a different, never observed environment (e.g. new climate)

  • Approach:

    • Randomized Controlled Trials extract causal mechanisms

    • Observational data is limited in quality: 

      • Large sampling bias

      • Few samples in contexts where given technique doesn’t work, many where it does: hard to identify the causal dynamics 

      • Unobserved heterogeneity

        • How farmers treat crops

        • Details of environment surrounding the farm 

    • Trials/Observational approach is limited: 

      • Cannot freely intervene on real systems

      • Dimensionality of real systems is too large

    • Proposed alternative:

      • Combination of expert, observational and modeled data

      • Hybrid structural models:

        • Process-based models (simulation of crop behavior, environment, soil, etc.)

        • Behavioral models (farmer decisions)

  • Prediction for future scenarios

    • Requires participatory process to envision future scenarios

      • Climate

      • Ag technologies and practices

    • Propose range of plausible future worlds

• Pathways forward: New Data Infrastructure for Agriculture

  • Many challenges in coordinating community

    • Data ownership

    • Voluntary vs Mandatory

    • Soft vs Hard Infrastructure

  • Alternative: remote sensing