From Farm to Planet: The InSEEDS World-Earth Model for Simulating Transitions to Regenerative Agriculture

Luana Schwarz, Jannes Breier, Hannah Prawitz, Max Bechthold, Werner von Bloh, Sara M. Constantino, Dieter Gerten, Jobst Heitzig, Ronja Hotz, Leander John, Christoph Müller, Johan Rockström, Jonathan F. Donges2025 · DOI

TL;DR

This paper introduces InSEEDS, a novel World-Earth model that integrates a biophysical vegetation model with an agent-based model of farmer decision-making. It simulates the complex social and ecological dynamics driving transitions to regenerative agriculture, offering a powerful tool to explore future land-use scenarios and inform policy for more resilient food systems.

What Problem Does This Paper Address?

Industrialized agriculture has led to significant environmental degradation and pushed planetary boundaries. While regenerative agriculture (RA) offers a promising alternative, current research often focuses on individual practices and their biophysical impacts, neglecting the complex social dynamics that drive their large-scale adoption. Factors like social-ecological feedbacks, tipping points, and transformative change remain underexplored. This gap hinders our ability to understand and promote the widespread transition to sustainable agricultural systems.

Methods

The core innovation is the InSEEDS (Integrated Social-Ecological resilient land Systems) model. It's a World-Earth Model (WEM) built on the copan:CORE framework, coupling the LPJmL Dynamic Global Vegetation Model (DGVM) with an agent-based model (ABM) of farmer decision-making.

  • LPJmL (ENV taxon): Simulates the biophysical impacts of agricultural practices, specifically focusing on conservation tillage (CT) and conventional farming (CF), and their effects on soil organic carbon (SOC) and crop yield (CY).
  • ABM (MET and CUL taxa): Represents farmers as agents with distinct types ("traditionalist" and "pioneer") who make decisions based on:
  • Attitudes (A): Formed by learning from their own land's performance and social learning from neighbors.
  • Subjective Norms (N): Influenced by the prevalent behaviors of neighboring farmers.
  • Perceived Behavioral Control (PBC): Reflects the perceived ease or difficulty of switching practices, with inertia introduced after a switch.
  • Theory of Planned Behavior (TPB): The ABM operationalizes farmer decision-making using TPB, integrating attitudes, norms, and PBC to predict behavioral intention and subsequent behavior (adoption of CF or CT).
  • Co-evolutionary Dynamics: The model explicitly links ecological outcomes (SOC, CY) to farmer decisions, creating feedback loops that drive co-evolutionary dynamics between human actions and the Earth system.

Datasets

The paper describes the use of existing datasets for model initialization:

  • Historical climate input data: From ISIMIP3a.
  • Land-use data, fertilizer, and manure inputs: Obtained from Ostberg et al. (2023).
  • Residue management data: Applied from Dietrich et al. (2020).
  • Tillage dataset: Created by Porwollik et al. (2019), providing global distribution of conservation agriculture practices.

Key Insights and Ideas

  • Social and Ecological Heterogeneity are Crucial: The model highlights that both the diversity of farmer types (AFTs) and the varying ecological conditions significantly influence adoption dynamics and outcomes.
  • Social Networks Matter: Local social norms and social learning from neighbors play a critical role in driving the adoption and spread of regenerative practices, sometimes even overriding purely ecological performance considerations.
  • Inertia is a Significant Barrier: The concept of Perceived Behavioral Control (PBC) and the inertia introduced after a management switch significantly inhibit rapid transitions, reflecting the real-world challenges of adopting new practices.
  • Context is King: The effectiveness and adoption of conservation tillage are highly context-dependent, varying significantly across different regions and their specific biophysical and social conditions.
  • Regenerative Agriculture's Potential is Underestimated: By focusing solely on conservation tillage as a representative RA practice, the model might present a conservative estimate of RA's potential positive ecological impacts.

Key Results

  • Parameter Sensitivity: Variations in attitude and norm weights have the strongest impact on simulation results, influencing the share of RA practitioners and resulting SOC/CY levels. Lowering social norm influence and increasing attitude influence leads to higher RA adoption and better ecological outcomes.
  • AFT Share Impact: The initial share of "pioneer" farmers significantly impacts adoption rates, with a 100% pioneer scenario showing a 50% higher adoption rate of CT compared to a 100% traditionalist scenario.
  • Global CT Benefits: On average, CT adoption leads to positive effects on SOC and CY, though these effects are less pronounced when social norms strongly favor conventional practices.
  • Regional Differences: Simulation results show distinct dynamics across countries. For instance, Paraguay exhibits a trend of SOC and CY loss due to legacy effects of land-use change, while Spain and South Africa show positive impacts of CT.
  • Norm-Dominated vs. Attitude-Dominated Switches: The model illustrates how switches to RA can be driven by either strong social norms (neighbor adoption) or by changing attitudes (learning from own land or social learning), with different implications for ecological performance.

Limitations and Future Work

  • Computational Demands: Running InSEEDS requires significant computational resources, limiting accessibility for extensive analyses like large ensembles or Monte Carlo simulations.
  • Simplified Farmer Representation: The current ABM has a relatively homogeneous representation of farmer decision-making, with limited inclusion of economic and political factors. Future work aims to refine individual decision-making, incorporate more learning processes, and introduce greater heterogeneity.
  • Limited RA Practices: The model currently focuses on conservation tillage as the primary RA practice. Future work should incorporate a wider range of RA practices (e.g., water management, intercropping, agroforestry) and consider context-specific tailoring.
  • Simplified Network Topology: The farmer network is currently limited to local, grid-cell-based connections (Moore Neighborhood). Future refinements will include multi-layer networks with different topologies to capture non-local connections.
  • Lack of Stochasticity in Simulations: While initial AFT allocation is stochastic, decision-making processes within simulations are deterministic. Future versions may incorporate stochasticity to better represent variability and uncertainty.
  • Need for Broader Agent Types: The current model focuses on farmers. Future work should incorporate other agent types (e.g., consumers, governments) and higher levels of social organization.
  • Empirical Validation: While the model is grounded in theory, further empirical validation with diverse datasets and regional behavioral data is crucial.

Summary updated

Abstract

Industrialised agriculture and its externalisation of environmental costs have contributed to accelerating ecological degradation and the transgression of planetary boundaries. Vice versa, agriculture is increasingly affected by ecological pressures such as climate change. While sustainable approaches like Regenerative Agriculture offer promising alternatives, most studies focus on the biophysical impacts of individual practices and overlook the complex dynamics underlying their large-scale adoption. In particular, the roles of social-ecological feedbacks, tipping dynamics, and transformative change remain underexplored. To address this gap, we introduce the InSEEDS integrated World-Earth model – a novel co-evolutionary approach to simulating agricultural transitions that couples a process-based vegetation model (LPJmL) with an agent-based model of farmer decision-making. InSEEDS integrates socio-cultural, social-ecological, and biophysical dynamics and can be applied from local to global scales. Distinguishing between a traditionalist and pioneer farmer types, we analyse the adoption dynamics of conservation tillage as a key practice of Regenerative Agriculture. We find that social networks, ecological heterogeneity, and decision-making inertia play a critical role in determining transition dynamics. Adoption of conservation tillage yields overall positive effects on soil carbon and crop yield, though outcomes are strongly context-dependent. InSEEDS provides a foundational tool that opens up avenues for understanding complex human-environment interactions in land-use transformations and advancing the next generation of World-Earth models.
BibTeX
@misc{schwarz2025from,
  author = {Luana Schwarz and Jannes Breier and Hannah Prawitz and Max Bechthold and Werner von Bloh and Sara M. Constantino and Dieter Gerten and Jobst Heitzig and Ronja Hotz and Leander John and Christoph Müller and Johan Rockström and Jonathan F. Donges},
  title = {{From Farm to Planet: The InSEEDS World-Earth Model for Simulating Transitions to Regenerative Agriculture}},
  year = {2025},
  doi = {10.5194/egusphere-2025-4079},
}
Figure