pith. sign in

arxiv: 2605.03648 · v1 · submitted 2026-05-05 · 💻 cs.AI

Agent-Based Modeling of Low-Emission Fertilizer Adoption for Dairy Farm Decarbonisation using Empirical Farm Data

Surya Jayakumar , Kieran Sullivan , John McLaughlin , Christine OMeara , Indrakshi Dey This is my paper

Pith reviewed 2026-05-07 16:27 UTC · model grok-4.3

classification 💻 cs.AI
keywords agent-based modelinglow-emission fertilizerdairy farm decarbonizationsocial contagionadoption dynamicspolicy simulationinnovation diffusiongreenhouse gas abatement
0
0 comments X

The pith

An agent-based model of 295 Irish dairy farms reproduces observed low-emission fertilizer adoption trajectories with R² of 0.979.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds an agent-based model to simulate how dairy farmers decide to switch to low-emission fertilizers over 15 years. It draws on real data from 295 Irish farms to include differences in farm size, social networks that spread ideas through peer influence, and policy levers such as subsidies or carbon taxes. The model matches actual adoption patterns closely and projects total greenhouse gas cuts plus the split between private and social costs. A reader would care because it turns decarbonization into a testable social-diffusion process instead of an abstract economic calculation, giving policymakers a virtual lab to try interventions first.

Core claim

The agent-based model, built on empirical data from 295 Irish dairy farms, represents farm communication via a social network and sets adoption probabilities from social contagion, farm-scale traits, and policy interventions. It reproduces observed adoption trajectories with R² = 0.979 and RMSE = 0.0274, and passes a Kolmogorov-Smirnov test against empirical data (D = 0.2407, p < 0.001). Adoption follows a logistic curve consistent with Rogers' diffusion theory and saturates near 91 percent. The framework then estimates sectoral emissions abatement and private-social cost trade-offs by treating decarbonization as a socio-technical diffusion process.

What carries the argument

The agent-based model with an explicit social network for peer influence and discussion groups, where each farm's adoption probability is driven by social contagion, farm characteristics, and policy inputs such as subsidies and carbon taxes.

If this is right

  • Subsidies and carbon taxes can be varied inside the model to quantify their separate effects on adoption speed and total emissions abatement.
  • The logistic diffusion pattern implies that early policy support can shift the whole sector toward saturation within roughly 15 years.
  • Monte Carlo runs allow planners to bound uncertainty in cumulative abatement and cost trade-offs before any real policy is launched.
  • The in silico laboratory can test the robustness of mitigation strategies to changes in network density or farm heterogeneity.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same social-contagion structure could be reused for modeling uptake of other farm practices such as methane inhibitors or cover cropping.
  • Strengthening discussion groups might substitute for part of the financial incentive needed to reach high adoption.
  • Adding external shocks such as feed-price spikes or new regulations would make the 15-year projections more realistic for real policy use.

Load-bearing premise

The social network structure, peer-influence strengths, and policy-response rules fitted to the 295-farm dataset will keep representing real decision-making across the full 15-year horizon without large unobserved changes or external shocks.

What would settle it

Long-term tracking of actual low-emission fertilizer adoption rates on Irish dairy farms over the next decade or more, checked against the model's projected cumulative curve and 91 percent saturation level.

Figures

Figures reproduced from arXiv: 2605.03648 by Christine OMeara, Indrakshi Dey, John McLaughlin, Kieran Sullivan, Surya Jayakumar.

Figure 1
Figure 1. Figure 1: Methodological framework 3.1 Dataset The study uses a farm-level dataset comprising 295 observations from multiple Irish farms, provided by Teagasc-the Agriculture and Food Development Authority of Ireland. The data was aggregated from two primary sources: the Irish Cattle Breeding Federation (ICBF), which supplied herd￾level performance data such as dairy cow numbers, calving patterns, milk yield, and mil… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the agent-based modelling framework. view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of top-performing and laggard farms in terms of carbon intensity and produc view at source ↗
Figure 4
Figure 4. Figure 4: Sensitivity of Carbon Farming Adoption to Varied Policy Instruments and Incentive Mag view at source ↗
Figure 5
Figure 5. Figure 5: Sensitivity Analysis of the Social Contagion Parameter ( view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of ABM-simulated adoption dynamics with the Rogers diffusion curve view at source ↗
Figure 7
Figure 7. Figure 7: Adoption diffusion (left) and projected annual fertilizer GHG emissions (right) under base view at source ↗
Figure 8
Figure 8. Figure 8: Adoption velocity across policy scenarios. view at source ↗
Figure 9
Figure 9. Figure 9: Cumulative GHG abatement over the 15-year simulation horizon. view at source ↗
Figure 10
Figure 10. Figure 10: Kernel density distribution of farm-level carbon intensity under baseline and subsidy view at source ↗
Figure 11
Figure 11. Figure 11: Model convergence diagnostics for cumulative abatement under the policy scenario ( view at source ↗
Figure 12
Figure 12. Figure 12: Longitudinal Network Snapshots of Spatial-Social Diffusion under a Subsidy Scenario view at source ↗
read the original abstract

To understand complex system dynamics in dairy farming, it is essential to use modeling tools that capture farm heterogeneity, social interactions, and cumulative environmental impacts. This study proposes an agent-based modeling (ABM) framework to simulate nitrogen management and the adoption of low-emission fertilizer across 295 Irish dairy farms over a 15-year period. Using empirical data, the model represents farm communication through a social network, capturing peer influence and discussion group dynamics, where adoption probabilities are driven by social contagion, farm-scale characteristics, and policy interventions such as subsidies and carbon taxes. The framework estimates sectoral greenhouse gas emissions, cumulative abatement, and private-social cost trade-offs, using Monte Carlo simulation and sensitivity analysis to quantify uncertainty. The model shows strong agreement with observed adoption trajectories ($R^2 = 0.979$, RMSE = 0.0274) and is validated against empirical data using a Kolmogorov-Smirnov test (D = 0.2407, p < 0.001), indicating its ability to reproduce structural patterns in adoption behavior. Adoption dynamics are further characterized using a logistic diffusion model consistent with Rogers' innovation diffusion theory, capturing progression from early adoption to a saturation level of approximately 91%. By framing decarbonization as a socio-technical diffusion process rather than a purely economic optimization problem, this study provides an in silico policy laboratory for evaluating the robustness and diffusion speed of climate mitigation strategies prior to implementation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript develops an agent-based model (ABM) of low-emission fertilizer adoption across 295 Irish dairy farms over a 15-year horizon. Farms are heterogeneous agents connected via an empirically derived social network; adoption probabilities depend on social contagion, farm-scale covariates, and policy levers (subsidies, carbon taxes). The model is calibrated to observed adoption trajectories, reports R² = 0.979 and RMSE = 0.0274, applies a Kolmogorov-Smirnov test (D = 0.2407, p < 0.001) for validation, fits a logistic diffusion curve, and uses Monte Carlo simulation to project sectoral GHG abatement and cost trade-offs.

Significance. If the validation evidence can be made internally consistent, the work supplies a policy-relevant in silico laboratory that couples empirical network data with diffusion dynamics and uncertainty quantification. The use of real farm-level data and explicit Monte Carlo propagation are strengths; the framing as a socio-technical diffusion process rather than pure optimization is a useful conceptual contribution.

major comments (2)
  1. [Abstract and §4] Abstract and §4 (validation): The reported two-sample Kolmogorov-Smirnov statistic (D = 0.2407, p < 0.001) with n ≈ 295 rejects the null of distributional equivalence at conventional significance levels (critical value ≈ 0.079 for α = 0.05). This directly contradicts the claim that the test “indicates its ability to reproduce structural patterns.” The manuscript must either (a) specify an alternative test quantity (e.g., residuals of the logistic fit or a one-sample procedure) or (b) revise the validation language and interpretation.
  2. [§3.2 and §4] §3.2 and §4: Adoption probabilities and the logistic diffusion parameters appear to be calibrated on the same 295-farm trajectories later used for the R² and KS comparisons. The manuscript should report an explicit out-of-sample or cross-validation protocol (or justify why the reported R² and KS constitute genuine predictive validation rather than in-sample fit).
minor comments (2)
  1. [§2.3] Notation for the social-contagion strength parameter and the policy-response coefficients should be defined once in a table or equation block rather than re-introduced in multiple sections.
  2. [Figures 3–5] Figure captions for the adoption trajectories and KS plots should state the exact sample sizes and whether the plotted distributions are cumulative or density.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and insightful comments on our manuscript. We address each of the major comments below and indicate the revisions we will make to strengthen the paper.

read point-by-point responses

  1. Referee: [Abstract and §4] Abstract and §4 (validation): The reported two-sample Kolmogorov-Smirnov statistic (D = 0.2407, p < 0.001) with n ≈ 295 rejects the null of distributional equivalence at conventional significance levels (critical value ≈ 0.079 for α = 0.05). This directly contradicts the claim that the test “indicates its ability to reproduce structural patterns.” The manuscript must either (a) specify an alternative test quantity (e.g., residuals of the logistic fit or a one-sample procedure) or (b) revise the validation language and interpretation.

    Authors: We agree with the referee that the interpretation of the Kolmogorov-Smirnov test requires revision. The low p-value indicates a statistically significant difference between the simulated and observed distributions, which does not support the claim of reproducing structural patterns. We will revise the language in the abstract and Section 4 to remove this claim and instead emphasize the high R² = 0.979 and low RMSE = 0.0274 as evidence of good agreement with observed trajectories. We will also consider adding a one-sample KS test against the fitted logistic curve or residual analysis as an alternative validation metric. This change will be incorporated in the revised manuscript. revision: yes

  2. Referee: [§3.2 and §4] §3.2 and §4: Adoption probabilities and the logistic diffusion parameters appear to be calibrated on the same 295-farm trajectories later used for the R² and KS comparisons. The manuscript should report an explicit out-of-sample or cross-validation protocol (or justify why the reported R² and KS constitute genuine predictive validation rather than in-sample fit).

    Authors: We acknowledge that the model parameters, including adoption probabilities and logistic diffusion parameters, were calibrated using the full set of 295 observed farm trajectories, making the R² and KS metrics in-sample goodness-of-fit measures rather than out-of-sample predictions. This is typical for agent-based models that integrate empirical data for network structure and covariates. To address this, we will add a clear statement in Section 4 explaining the calibration-validation approach and justify that the reproduction of observed adoption trajectories serves as validation for the socio-technical diffusion process. Additionally, we will explore implementing a cross-validation protocol by partitioning the farms (e.g., by region or size) or using Monte Carlo cross-validation to assess robustness, and report this in the revised version if feasible. If data limitations prevent a full out-of-sample test, we will discuss this as a limitation of the study. revision: partial

Circularity Check

1 steps flagged

R² agreement and KS validation reduce to fit on the same empirical trajectories used for calibration

specific steps
  1. fitted input called prediction [Abstract]
    "The model shows strong agreement with observed adoption trajectories (R² = 0.979, RMSE = 0.0274) and is validated against empirical data using a Kolmogorov-Smirnov test (D = 0.2407, p < 0.001), indicating its ability to reproduce structural patterns in adoption behavior. Adoption dynamics are further characterized using a logistic diffusion model consistent with Rogers' innovation diffusion theory, capturing progression from early adoption to a saturation level of approximately 91%."

    Adoption probabilities and diffusion parameters are set using the 295-farm empirical dataset; the R² fit, KS comparison, and logistic characterization are then evaluated against the observed adoption trajectories from that same dataset, rendering the reported agreement a calibration diagnostic rather than an independent test.

full rationale

The ABM parameters for peer influence, social contagion, and adoption probabilities are derived directly from the 295-farm empirical dataset. The reported R²=0.979 agreement with observed adoption trajectories and the KS test are then performed against adoption patterns from that identical dataset. This makes the 'strong agreement' and 'reproduce structural patterns' claims measures of in-sample fit quality rather than out-of-sample prediction or independent validation. The logistic diffusion characterization is likewise fitted to the same trajectories. While the model structure itself (network + policy rules) adds content, the central quantitative validation claim reduces to a fitted-input-called-prediction pattern.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

Based on abstract alone, the model rests on standard assumptions about social contagion and logistic diffusion but introduces several fitted parameters for network influence and policy response whose values are not independently derived.

free parameters (2)
  • social contagion strength
    Adoption probabilities are driven by peer influence parameters calibrated to match observed trajectories.
  • policy response coefficients
    Effects of subsidies and carbon taxes are parameterized to reproduce empirical adoption patterns.
axioms (2)
  • domain assumption Farmers' adoption decisions follow a logistic diffusion process consistent with Rogers' innovation diffusion theory.
    Invoked to characterize progression from early adoption to saturation.
  • domain assumption Social network structure derived from discussion groups accurately captures peer influence.
    Central to the agent communication mechanism.

pith-pipeline@v0.9.0 · 5569 in / 1425 out tokens · 69371 ms · 2026-05-07T16:27:42.574960+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

  1. [1]

    [EPA(2025)] EPA. (2025). Dairy farming is a major contributor to greenhouse gas emissions in Ireland. Environmental Protection Agency, Ireland. [Online]. Available: https://www.epa.ie/ [Teagasc(2024)] Teagasc. (2024). National Farm Survey 2023: Sustainability and Efficiency Indi- cators for Irish Dairy Farms. Teagasc, Athenry, Ireland. [Teagasc (2024)] Te...

    work page 2025

  2. [2]

    https://doi.org/10.1186/s40100-024-00306- w [Ferrer et al.(2023)] Ferrer, A. J. G., Thanh, L. H., & Chuong, P. H. (2023). Farming household adoption of climate-smart agricultural technologies: evidence from North-Central Vietnam.Cli- mate and Development, 7(2), 641–663. https://doi.org/10.1007/s41685-023-00296-5 27 [Jameson et al.(2024)] Jameson, J., McDo...

    work page doi:10.1186/s40100-024-00306- 2023

  3. [3]

    R., & Renwick, A

    https://doi.org/10.3390/su16041419 [Eastwood & Renwick(2020)] Eastwood, C. R., & Renwick, A. (2020). Innovation uncertainty im- pacts the adoption of smarter farming approaches.Frontiers in Sustainable Food Systems,

    work page doi:10.3390/su16041419 2020

  4. [4]

    https://doi.org/10.3389/fsufs.2020.00024 [Chapman et al.(2022)] Chapman, M., Wiltshire, S., Baur, P., et al. (2022). Social-ecological feedbacks drive tipping points in farming system diversification. arXiv. https://arxiv.org/abs/2205.04425 [Alotibi(2025)] Alotibi, Y. S. (2025). A socio-technical agent-based simulation model for predicting smart agricultu...

    work page doi:10.3389/fsufs.2020.00024 2020

  5. [5]

    [Sinnott et al.(2025)] Sinnott, U., Hanrahan, K., & Wallace, M. (2025). Identifying farmer ty- pologies based on attitudes towards greenhouse gas mitigation in Irish agriculture.Journal of Agricultural Education and Extension. [Palma-Molina et al.(2023)] Palma-Molina, P., Shalloo, M., & Horan, B. (2023). Factors associated with the intensity of technology...

    work page 2025

  6. [6]

    [Serebrennikov et al.(2020)] Serebrennikov, D., Thorne, F., Kallas, Z., & McCarthy, S. N. (2020). Factors influencing adoption of sustainable farming practices in Europe: A systemic review of empirical literature.Sustainability, 12(22),

    work page 2020

  7. [7]

    https://doi.org/10.3390/su12229719 [Feng et al.(2023)] Feng, J., Li, H., Cannon, N., Chang, X., & Chu, Q. (2023). A simulation study of how Chinese farmer cooperatives can drive effective low-carbon production systems through a carbon transaction incentive. [Marvuglia et al.(2021)] Marvuglia, A., Bayram, A., Baustert, P., Navarrete Guti´ errez, T., & Igos...

    work page doi:10.3390/su12229719 2023

  8. [8]

    https://doi.org/10.18564/jasss.4714 [Bayram et al.(2024)] Bayram, A., Marvuglia, A., Navarrete Guti´ errez, T., & Soyeurt, H. (2024). Balancing environmental sustainability and economic viability in Luxembourgish farms: An agent-based model with multi-objective optimization.Sustainability, 16,

    work page doi:10.18564/jasss.4714 2024

  9. [9]

    https://doi.org/10.3390/su16198536 [Bayram (2023)] Bayram, A., Marvuglia, A., Navarrete Gutierrez, T., Weis, J.-P., Conter, G., & Zimmer, S. (2023). Sustainable farming strategies for mixed crop-livestock farms in Luxem- bourg simulated with a hybrid agent-based and life-cycle assessment model.Journal of Cleaner Production, 386, 135759. https://doi.org/10...

    work page doi:10.3390/su16198536 2023

  10. [10]

    Reducing nitrous ox- ide emissions by changing N fertiliser use from calcium ammonium nitrate (CAN) to urea based formulations,

    [Online]. Available: https://teagasc.ie/wp-content/uploads/media/website/publications/2024/National-Farm- Survey-2023.pdf [Citizens Information (2026)] Citizens Information (2026). Carbon tax. Available at:https: //www.citizensinformation.ie/en/money-and-tax/tax/motor-carbon-other-taxes/ carbon-tax/ [Teagasc (2023)] Teagasc (2023). Protected urea: Reducin...

    work page doi:10.1016/j.scitotenv.2016.04.120 2024

  11. [11]

    Effects of extreme events on land-use-related decisions of farmers in Eastern Austria: the role of learning,

    [Online]. Available:https://www.teagasc.ie/news/ protected-urea-performance/. [Accessed: Jan. 8, 2026]. [Williams (2025)] Williams, T. G., Brown, C., Diogo, V., Magliocca, N. R., Molla, N., Rounsevell, M. D. A., Zagaria, C., & Verburg, P. H. (2025). Power dynamics shape sustainability transitions in a modeled food system.One Earth, 8(1), 101158. https://d...

    work page doi:10.1016/j.oneear.2024.11.012 2026

  12. [12]

    Assessing the impacts of climate change on agriculture and water systems via coupled human-hydrological modeling,

    [Javansalehi(2024)] M. Javansalehi and M. Shourian, “Assessing the impacts of climate change on agriculture and water systems via coupled human-hydrological modeling,”Agricultural Water Management,

    work page 2024

  13. [13]

    The uptake of European soy production: Insights from a partic- ipatory agent-based model,

    [Farahbakhsh(2025)] S. Farahbakhsh, K. Peeters, J. Pannecoucque, G. Tavernier, B. Vanden Berghen, and J. van Meensel, “The uptake of European soy production: Insights from a partic- ipatory agent-based model,”Agricultural Systems,

    work page 2025

  14. [14]

    [Tarruella (2025)] Tarruella, M., Huber, R., El-Benni, N., Mack, G., Sch¨ afer, D., & Finger, R. (2025). Understanding farmer behaviour for successful climate change mitigation in voluntary initiatives.Australian Journal of Agricultural and Resource Economics. 29

    work page 2025