pith. sign in

arxiv: 2604.04098 · v1 · submitted 2026-04-05 · 💻 cs.CV

A Physics-Informed, Behavior-Aware Digital Twin for Robust Multimodal Forecasting of Core Body Temperature in Precision Livestock Farming

Riasad Alvi , Mohaimenul Azam Khan Raiaan , Sadia Sultana Chowa , Arefin Ittesafun Abian , Reem E Mohamed , Md Rafiqul Islam , Yakub Sebastian , Sheikh Izzal Azid
show 1 more author
Sami Azam
This is my paper

Pith reviewed 2026-05-13 17:14 UTC · model grok-4.3

classification 💻 cs.CV
keywords digital twincore body temperature forecastingprecision livestock farmingphysics-informed modelingbehavioral Markov chainheat stress predictionmultimodal fusionstacked ensemble
0
0 comments X

The pith

A digital twin merges ODE thermoregulation equations with a behavioral Markov chain to forecast dairy cattle core body temperature two hours ahead.

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

The paper builds a physics-informed digital twin that runs an ordinary differential equation model of metabolic heat production and loss, corrects it with a Gaussian process for individual cow differences, aligns outputs to live sensors via Kalman filtering, and layers on a Markov chain to track shifts between activity states. These outputs are combined with raw multimodal sensor streams through multi-scale feature engineering and fed into a three-stage stacked LightGBM ensemble that produces the final temperature forecast plus uncertainty bounds. The resulting system is evaluated on the high-frequency MmCows dataset and reaches a cross-validated R-squared of 0.783, F1 of 84.25 percent, and prediction-interval coverage of 92.38 percent at the two-hour horizon. A reader would care because timely, uncertainty-aware heat-stress alerts could let farmers intervene earlier to protect animal welfare and maintain milk production.

Core claim

The central claim is that embedding an ODE-based thermoregulation model, a Gaussian process for cow-specific residuals, a Kalman filter for sensor fusion, and a behavioral Markov chain for activity-state transitions inside a digital twin produces physically consistent features that, when fused with heterogeneous sensor data in an expert-weighted stacked ensemble, deliver reliable two-hour-ahead core-body-temperature forecasts together with calibrated uncertainty intervals.

What carries the argument

The physics-informed digital twin that couples an ODE model of heat production and dissipation with a behavioral Markov chain to generate physiological and activity-state predictions for downstream fusion.

If this is right

  • Early CBT forecasts enable proactive heat-stress mitigation before visible symptoms appear.
  • Ablation results show that DT-derived features measurably lift ensemble accuracy over sensor-only baselines.
  • Bootstrapped uncertainty estimates allow risk-aware thresholds for farm alerts.
  • The three-stage ensemble structure separates modality-specific learning from final fusion, supporting incremental addition of new sensor types.

Where Pith is reading between the lines

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

  • The same digital-twin structure could be re-parameterized for other species by swapping species-specific heat-balance coefficients.
  • Embedding the twin inside an automated ventilation or cooling control loop would turn forecasts into direct actuation commands.
  • Extending the Markov chain to include feeding or milking events might improve longer-horizon predictions.
  • The framework's emphasis on physical consistency may reduce the data volume needed to retrain when cows move to new environments.

Load-bearing premise

The ODE thermoregulation equations plus the activity-state Markov chain capture the dominant physiological and behavioral dynamics present in the MmCows dataset without large unmodeled effects or farm-specific biases.

What would settle it

Running the same pipeline on an independent dataset collected from different breeds, climates, or housing systems and obtaining R-squared below 0.6 or PICP below 80 percent for the two-hour forecasts would falsify the robustness claim.

read the original abstract

Precision livestock farming requires accurate and timely heat stress prediction to ensure animal welfare and optimize farm management. This study presents a physics-informed digital twin (DT) framework combined with an uncertainty-aware, expert-weighted stacked ensemble for multimodal forecasting of Core Body Temperature (CBT) in dairy cattle. Using the high-frequency, heterogeneous MmCows dataset, the DT integrates an ordinary differential equation (ODE)-based thermoregulation model that simulates metabolic heat production and dissipation, a Gaussian process for capturing cow-specific deviations, a Kalman filter for aligning predictions with real-time sensor data, and a behavioral Markov chain that models activity-state transitions under varying environmental conditions. The DT outputs key physiological indicators, such as predicted CBT, heat stress probability, and behavioral state distributions are fused with raw sensor data and enriched through multi-scale temporal analysis and cross-modal feature engineering to form a comprehensive feature set. The predictive methodology is designed in a three-stage stacked ensemble, where stage 1 trains modality-specific LightGBM 'expert' models on distinct feature groups, stage 2 collects their predictions as meta-features, and at stage 3 Optuna-tuned LightGBM meta-model yields the final CBT forecast. Predictive uncertainty is quantified via bootstrapping and validated using Prediction Interval Coverage Probability (PICP). Ablation analysis confirms that incorporating DT-derived features and multimodal fusion substantially enhances performance. The proposed framework achieves a cross-validated R2 of 0.783, F1 score of 84.25% and PICP of 92.38% for 2-hour ahead forecasting, providing a robust, uncertainty-aware, and physically principled system for early heat stress detection and precision livestock management.

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

3 major / 2 minor

Summary. The manuscript proposes a physics-informed digital twin (DT) framework for 2-hour-ahead core body temperature (CBT) forecasting in dairy cattle. It integrates an ODE-based thermoregulation model, Gaussian process for cow-specific deviations, Kalman filter for sensor alignment, and behavioral Markov chain, then fuses DT outputs (predicted CBT, heat-stress probability, behavioral distributions) with raw sensor data via multi-scale temporal and cross-modal feature engineering. These features feed a three-stage stacked ensemble of modality-specific LightGBM experts whose predictions are combined by an Optuna-tuned LightGBM meta-model. Cross-validated results on the MmCows dataset are reported as R² = 0.783, F1 = 84.25 %, and PICP = 92.38 %, with ablation studies attributing gains to the DT-derived features.

Significance. If the central claims hold after the requested validation, the work would constitute a useful contribution to precision livestock farming by demonstrating how standard thermoregulation ODEs and Markov activity models can be turned into informative features for an uncertainty-aware ensemble. The explicit ablation, multimodal fusion, and PICP reporting are positive elements that distinguish it from purely data-driven baselines.

major comments (3)
  1. [Abstract and §4] Abstract and §4 (results): the headline metrics (R² = 0.783, F1 = 84.25 %, PICP = 92.38 %) are obtained only after the DT-derived features are added; however, no quantitative fidelity check is supplied for the ODE thermoregulation model or the Markov chain on the MmCows data (e.g., residual norm on held-out cows, Markov log-likelihood on observed state sequences, or sensitivity of final metrics to ODE-parameter perturbation). Without such checks the DT features could be correlated noise rather than physiologically grounded signal.
  2. [Abstract and §3] Abstract and §3 (methods): the manuscript states that cross-validation was performed but supplies no information on the data-splitting strategy, the exact provenance of the ODE parameters, or the precise post-hoc feature-engineering steps. These omissions make it impossible to verify that the reported performance is free of leakage or dataset-specific overfitting.
  3. [Abstract] Abstract: the PICP of 92.38 % is presented as validation of uncertainty quantification, yet it is unclear whether the interval construction accounts for uncertainty arising from the ODE parameters, the Markov transition matrix, or the Gaussian-process deviations; if these sources are omitted the coverage probability may be over-optimistic.
minor comments (2)
  1. [Abstract and §3.3] The abstract describes the ensemble as 'expert-weighted' while §3.3 only mentions an Optuna-tuned LightGBM meta-model; the weighting mechanism (if any) should be stated explicitly.
  2. Notation for the DT outputs (heat-stress probability, behavioral-state distributions) is introduced without accompanying equations; adding a compact mathematical definition would improve clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below and will revise the manuscript to improve clarity, reproducibility, and validation of the physics-informed components.

read point-by-point responses

  1. Referee: [Abstract and §4] Abstract and §4 (results): the headline metrics (R² = 0.783, F1 = 84.25 %, PICP = 92.38 %) are obtained only after the DT-derived features are added; however, no quantitative fidelity check is supplied for the ODE thermoregulation model or the Markov chain on the MmCows data (e.g., residual norm on held-out cows, Markov log-likelihood on observed state sequences, or sensitivity of final metrics to ODE-parameter perturbation). Without such checks the DT features could be correlated noise rather than physiologically grounded signal.

    Authors: We agree that explicit fidelity checks for the standalone ODE thermoregulation model and Markov chain would strengthen the interpretation that DT-derived features provide physiologically grounded signal. In the revised manuscript we will add a dedicated subsection in §4 reporting: (i) residual norms between ODE-predicted CBT and observed values on held-out cows, (ii) log-likelihood of the fitted Markov chain on observed behavioral state sequences, and (iii) a sensitivity analysis showing the effect of ODE-parameter perturbations on the final ensemble R² and PICP. These additions will directly address the concern that the features might be correlated noise. revision: yes

  2. Referee: [Abstract and §3] Abstract and §3 (methods): the manuscript states that cross-validation was performed but supplies no information on the data-splitting strategy, the exact provenance of the ODE parameters, or the precise post-hoc feature-engineering steps. These omissions make it impossible to verify that the reported performance is free of leakage or dataset-specific overfitting.

    Authors: We acknowledge the need for full methodological transparency. The revised §3 will explicitly describe: the cross-validation strategy (cow-stratified time-series split to avoid temporal leakage), the provenance of ODE parameters (literature-derived values from thermoregulation studies, with cow-specific deviations captured by the Gaussian process), and the exact sequence of multi-scale temporal and cross-modal feature-engineering operations. These details will enable independent verification and confirm absence of leakage. revision: yes

  3. Referee: [Abstract] Abstract: the PICP of 92.38 % is presented as validation of uncertainty quantification, yet it is unclear whether the interval construction accounts for uncertainty arising from the ODE parameters, the Markov transition matrix, or the Gaussian-process deviations; if these sources are omitted the coverage probability may be over-optimistic.

    Authors: The current PICP is computed from bootstrapped ensemble predictions, which capture model variability but do not explicitly propagate uncertainty from the DT components. In the revision we will clarify this limitation in the abstract and methods, and we will add a Monte-Carlo propagation of ODE-parameter and GP-deviation uncertainty into the prediction intervals where computationally feasible. Updated PICP values will be reported if the coverage changes materially. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation chain is self-contained

full rationale

The paper constructs a digital twin from standard ODE thermoregulation equations (not self-defined), a Gaussian process for deviations, Kalman filtering for sensor alignment, and a Markov chain for behavior. These DT outputs are then used as additional features fed into a separate stacked LightGBM ensemble that is trained to forecast observed CBT values. No equation reduces the final forecast directly to a fitted parameter or DT output by construction; the ensemble learns mappings from the fused feature set to ground-truth CBT on cross-validation splits. No load-bearing self-citations or uniqueness theorems from prior author work are invoked. The reported R2, F1, and PICP therefore reflect empirical performance of the ensemble rather than tautological reuse of inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The framework rests on standard mathematical assumptions for ODE integration and Kalman filtering plus domain assumptions about cow thermoregulation and activity-state transitions; no new physical constants or entities are introduced beyond the digital-twin construct itself.

free parameters (1)
  • Optuna-tuned LightGBM hyperparameters
    Stage-3 meta-model hyperparameters are optimized on the training data and therefore constitute fitted parameters that affect the reported R².
axioms (2)
  • domain assumption The ODE thermoregulation model accurately represents metabolic heat production and dissipation under the environmental conditions present in the MmCows dataset.
    Invoked when the digital twin is described as integrating the ODE-based model to simulate core body temperature dynamics.
  • domain assumption The behavioral Markov chain correctly models activity-state transitions as a function of environmental variables.
    Used to generate behavioral state distributions that are later fused into the feature set.

pith-pipeline@v0.9.0 · 5654 in / 1525 out tokens · 26542 ms · 2026-05-13T17:14:32.903251+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

68 extracted references · 68 canonical work pages

  1. [1]

    A review of thermal stress in cattle,

    R. Shephard and S. Maloney, “A review of thermal stress in cattle,”Australian veterinary journal, vol. 101, no. 11, pp. 417–429, 2023

    work page 2023

  2. [2]

    Invited review: Effects of heat stress on dairy cattle welfare,

    L. Polsky and M. A. Von Keyserlingk, “Invited review: Effects of heat stress on dairy cattle welfare,”Journal of dairy science, vol. 100, no. 11, pp. 8645–8657, 2017

    work page 2017

  3. [3]

    Heat stress effects on livestock: molecular, cellular and metabolic aspects, a review,

    I. B. Slimen, T. Najar, A. Ghram, and M. Abdrrabba, “Heat stress effects on livestock: molecular, cellular and metabolic aspects, a review,”Journal of Animal Physi- ology and Animal Nutrition, vol. 100, no. 3, pp. 401–412, 2016

    work page 2016

  4. [4]

    The effects of heat stress on the behaviour of dairy cows–a review,

    P. Herbut, G. Hoffmann, S. Angrecka, D. Gody´ n, F. M. C. Vieira, K. Adamczyk, and R. Kupczy´ nski, “The effects of heat stress on the behaviour of dairy cows–a review,”Annals of Animal Science, vol. 21, no. 2, pp. 385–402, 2021

    work page 2021

  5. [5]

    Increasing impor- tance of heat stress for cattle farming under future global climate scenarios,

    M. A. Carvajal, A. J. Alaniz, C. Guti´ errez-G´ omez, P. M. Vergara, V. Sejian, and F. Bozinovic, “Increasing impor- tance of heat stress for cattle farming under future global climate scenarios,”Science of The Total Environment, vol. 801, p. 149661, 2021

    work page 2021

  6. [6]

    Impact of heat stress on cattle systems: Responses of production metrics to thermal stress,

    F. M. Hasan, A. Chlingaryan, P. C. Thomson, C. E. F. Clark, M. R. Islam, and S. Lomax, “Impact of heat stress on cattle systems: Responses of production metrics to thermal stress,”Computers and Electronics in Agricul- ture, vol. 240, p. 111143, 2026

    work page 2026

  7. [7]

    Heat stress amelioration for pasture-based dairy cattle: challenges and opportunities,

    S. J. Hendriks, J. P. Edwards, A. K. Shirley, C. E. F. Clark, K. E. Sch¨ utz, K. J. Verhoek, and J. G. Jago, “Heat stress amelioration for pasture-based dairy cattle: challenges and opportunities,”Animal Frontiers, vol. 15, no. 2, pp. 32–42, Apr. 2025

    work page 2025

  8. [8]

    Effects of heat stress on endocrine, thermoregulatory, and lac- tation capacity in heat-tolerant and -sensitive dry cows,

    X. Chen, C. Li, T. Fang, J. Yao, and X. Gu, “Effects of heat stress on endocrine, thermoregulatory, and lac- tation capacity in heat-tolerant and -sensitive dry cows,” Frontiers in Veterinary Science, vol. 11, 2024

    work page 2024

  9. [9]

    Heat stress as a barrier to successful re- production and potential alleviation strategies in cattle,

    I. Khan, A. Mesalam, Y. S. Heo, S.-H. Lee, G. Nabi, and I.-K. Kong, “Heat stress as a barrier to successful re- production and potential alleviation strategies in cattle,” Animals, vol. 13, 2023

    work page 2023

  10. [10]

    Navigating the labyrinth of heat stress assessment in dairy cattle: A comprehensive review of methods and emerging tech- nologies,

    A. Singaravadivelan, A. Prasad, C. Balusami, S. Hariku- mar, B. Vakayil, V. L. Gleeja, V. Sejian, P. Vijayaku- mar, S. P. Bakkesh, I. P. Kumaret al., “Navigating the labyrinth of heat stress assessment in dairy cattle: A comprehensive review of methods and emerging tech- nologies,”Computers and Electronics in Agriculture, vol. 237, p. 110517, 2025

    work page 2025

  11. [11]

    Recent ad- vances on early detection of heat strain in dairy cows us- ing animal-based indicators: A review,

    H. Shu, W. Wang, L. Guo, and J. Bindelle, “Recent ad- vances on early detection of heat strain in dairy cows us- ing animal-based indicators: A review,”Animals, vol. 11, no. 4, 2021

    work page 2021

  12. [12]

    A review of mul- timodal ai in veterinary diagnosis: Current trends, chal- lenges, and future directions,

    C. Gomes, L. Coheur, and P. Tilley, “A review of mul- timodal ai in veterinary diagnosis: Current trends, chal- lenges, and future directions,”IEEE Access, 2025

    work page 2025

  13. [13]

    Internet of things (iot): Sensors application in dairy cattle farming,

    F. M. Tangorra, E. Buoio, A. Calcante, A. Bassi, and A. Costa, “Internet of things (iot): Sensors application in dairy cattle farming,”Animals, vol. 14, no. 21, p. 3071, 2024

    work page 2024

  14. [14]

    Predictive models of dairy cow ther- mal state: A review from a technological perspective,

    S. F. Neves, M. C. F. Silva, J. M. Miranda, G. Stilwell, and P. P. Cortez, “Predictive models of dairy cow ther- mal state: A review from a technological perspective,” Veterinary Sciences, vol. 9, no. 8, 2022

    work page 2022

  15. [15]

    Predictive models of dairy cow thermal state: a review from a technological perspective,

    S. F. Neves, M. C. Silva, J. M. Miranda, G. Stilwell, and P. P. Cortez, “Predictive models of dairy cow thermal state: a review from a technological perspective,”Vet- erinary sciences, vol. 9, no. 8, p. 416, 2022

    work page 2022

  16. [16]

    A review and future trends of precision livestock over dairy and beef cow cattle with artificial intelligence,

    ´A. Michelena, ´O. Fontenla-Romero, and J. Luis Calvo- Rolle, “A review and future trends of precision livestock over dairy and beef cow cattle with artificial intelligence,” Logic Journal of the IGPL, vol. 33, no. 4, p. jzae111, 2025

    work page 2025

  17. [17]

    Ther- mal imaging combined with predictive machine learning based model for the development of thermal stress level classifiers,

    V. M. Pacheco, R. V. de Sousa, A. V. da Silva Rodrigues, E. J. de Souza Sardinha, and L. S. Martello, “Ther- mal imaging combined with predictive machine learning based model for the development of thermal stress level classifiers,”Livestock Science, vol. 241, 2020

    work page 2020

  18. [18]

    Predictive models for heat stress as- sessment in holstein dairy heifers using infrared thermog- raphy and machine learning,

    A. L. V. Pereira, L. S. Martello, J. C. D. Campos, A. V. da Silva Rodrigues, G. P. de Carvalho Nunes Oliveira, and R. V. de Sousa, “Predictive models for heat stress as- sessment in holstein dairy heifers using infrared thermog- raphy and machine learning,”Tropical Animal Health and Production, vol. 56, no. 8, p. 285, 2024

    work page 2024

  19. [19]

    Ma- chine learning applications for precision agriculture: A comprehensive review,

    A. Sharma, A. Jain, P. Gupta, and V. Chowdary, “Ma- chine learning applications for precision agriculture: A comprehensive review,”IEEE Access, vol. 9, pp. 4843– 4873, 2020

    work page 2020

  20. [20]

    Internet of things-based health surveillance systems for livestock: A review of recent advances and challenges,

    M. Modak, M. M. Pritom, S. C. Banik, and M. S. Rabbi, “Internet of things-based health surveillance systems for livestock: A review of recent advances and challenges,” IET Wireless Sensor Systems, vol. 15, no. 1, p. e70013, 2025

    work page 2025

  21. [21]

    A systematic literature review on deep learning applications for precision cattle farming,

    M. S. Mahmud, A. Zahid, A. K. Das, M. Muzammil, and M. U. Khan, “A systematic literature review on deep learning applications for precision cattle farming,” Computers and Electronics in Agriculture, vol. 187, p. 106313, 2021

    work page 2021

  22. [22]

    Heat stress amelioration for pasture-based dairy cattle: challenges and opportunities,

    S. J. Hendriks, J. P. Edwards, A. K. Shirley, C. E. Clark, K. E. Sch¨ utz, K. J. Verhoek, and J. G. Jago, “Heat stress amelioration for pasture-based dairy cattle: challenges and opportunities,”Animal Frontiers, vol. 15, no. 2, pp. 32–42, 2025

    work page 2025

  23. [23]

    Heat stress in dairy cows: impacts, identifi- cation, and mitigation strategies—a review,

    C. P. Oliveira, F. C. d. Sousa, A. L. d. Silva, ´E. B. Schultz, R. I. Valderrama Londo˜ no, and P. A. R. d. 21 Souza, “Heat stress in dairy cows: impacts, identifi- cation, and mitigation strategies—a review,”Animals, vol. 15, no. 2, p. 249, 2025

    work page 2025

  24. [24]

    Optimized machine learning models for predicting core body temperature in dairy cows: Enhancing accuracy and interpretability for practical livestock management,

    D. Li, G. Yan, F. Li, H. Lin, H. Jiao, H. Han, and W. Liu, “Optimized machine learning models for predicting core body temperature in dairy cows: Enhancing accuracy and interpretability for practical livestock management,” Animals, vol. 14, no. 18, 2024

    work page 2024

  25. [25]

    Sensor- type agnostic heat detection in dairy cows using multi- autoencoders with shared latent space,

    D.-H. Lee, M. Lee, W.-H. Lee, and S. Seo, “Sensor- type agnostic heat detection in dairy cows using multi- autoencoders with shared latent space,”Applied Soft Computing, vol. 166, p. 112200, 2024

    work page 2024

  26. [26]

    A deep learning model to forecast cattle heat stress,

    N. H. Chapman, A. Chlingaryan, P. C. Thomson, S. Lo- max, M. A. Islam, A. K. Doughty, and C. E. F. Clark, “A deep learning model to forecast cattle heat stress,” Computers and Electronics in Agriculture, vol. 211, p. 107932, 2023

    work page 2023

  27. [27]

    Predicting dairy cattle heat stress using machine learning techniques,

    C. A. Becker, A. Aghalari, M. Marufuzzaman, and A. E. Stone, “Predicting dairy cattle heat stress using machine learning techniques,”Journal of Dairy Science, vol. 104, no. 1, pp. 501–524, 2021

    work page 2021

  28. [29]

    Identifying and predicting heat stress events for grazing dairy cows using rumen temperature boluses,

    S. J. R. Woodward, J. P. Edwards, K. J. Verhoek, and J. G. Jago, “Identifying and predicting heat stress events for grazing dairy cows using rumen temperature boluses,” JDS Communications, vol. 5, no. 5, pp. 431–435, 2024

    work page 2024

  29. [30]

    Detect- ing heat stress: Examination of temperature-humidity in- dex thresholds for respiration rate and body temperature in barn and pasture-housed peripubertal dairy heifers,

    K. Daniels, M. Ellett, C. Parsons, and B. Corl, “Detect- ing heat stress: Examination of temperature-humidity in- dex thresholds for respiration rate and body temperature in barn and pasture-housed peripubertal dairy heifers,” JDS communications, 2025

    work page 2025

  30. [31]

    Bimodal data analysis for early detection of lameness in dairy cows using artificial intelligence,

    Y. Dhaliwal, H. Bi, and S. Neethirajan, “Bimodal data analysis for early detection of lameness in dairy cows using artificial intelligence,”Journal of Agriculture and Food Research, vol. 21, p. 101837, 2025

    work page 2025

  31. [32]

    Cattle welfare assessment based on adaptive fuzzy logic and multimodal data fusion,

    L. Tong, J. Fang, X. Wang, and Y. Zhao, “Cattle welfare assessment based on adaptive fuzzy logic and multimodal data fusion,”Frontiers in Veterinary Science, vol. 12, 2025

    work page 2025

  32. [33]

    Computer vision systems for monitoring hutch-housed dairy calves,

    A. Negreiro, T. Bresolin, R. E. P. Ferreira, B. Dado- Senn, J. M. C. Van Os, J. Laporta, and J. R. R. D´ orea, “Computer vision systems for monitoring hutch-housed dairy calves,”Journal of Dairy Science, vol. 108, no. 9, pp. 9998–10 011, 2025

    work page 2025

  33. [34]

    Non- invasive assessment of heat comfort in dairy calves based on thermal signature,

    R. V. de Sousa, J. C. D. Campos, G. Pagin, D. F. Pereira, A. R. Concei¸ c˜ ao, R. A. Tabile, and L. S. Martello, “Non- invasive assessment of heat comfort in dairy calves based on thermal signature,”Dairy, vol. 6, no. 4, 2025

    work page 2025

  34. [35]

    Insights into thermal stress effects on performance and behavior of grazing cattle via multimodal sensor monitoring,

    R. Eckhardt, R. Arablouei, K. McCoskeret al., “Insights into thermal stress effects on performance and behavior of grazing cattle via multimodal sensor monitoring,”Sci- entific Reports, vol. 15, p. 27941, 2025

    work page 2025

  35. [36]

    Rumination time, reticulorumen temperature, and activity in relation to postpartum health status in dairy cows during heat stress,

    S. Szalai, ´A. Bodnar, H. Febel, M. Bakony, and V. Ju- rkovich, “Rumination time, reticulorumen temperature, and activity in relation to postpartum health status in dairy cows during heat stress,”Animals, vol. 15, no. 11, 2025

    work page 2025

  36. [37]

    The impact of heat stress on dairy cattle: Effects on milk quality, rumination behaviour, and reticulorumen ph response using machine learning models,

    K. Dˇ zermeikait˙ e, J. Kriˇ stolaityt˙ e, D. Malaˇ sauskien˙ e, S. Arlauskait˙ e, A. Girdauskait˙ e, and R. Antanaitis, “The impact of heat stress on dairy cattle: Effects on milk quality, rumination behaviour, and reticulorumen ph response using machine learning models,”Biosensors, vol. 15, no. 9, 2025

    work page 2025

  37. [38]

    Predicting dairy cattle heat stress using machine learn- ing techniques,

    C. Becker, A. Aghalari, M. Marufuzzaman, and A. Stone, “Predicting dairy cattle heat stress using machine learn- ing techniques,”Journal of dairy science, vol. 104, no. 1, pp. 501–524, 2021

    work page 2021

  38. [39]

    Global projections of heat stress at high temporal resolution using machine learn- ing,

    P. Georgiades, T. Economou, Y. Proestos, J. Araya, J. Lelieveld, and M. Neira, “Global projections of heat stress at high temporal resolution using machine learn- ing,”Earth System Science Data, vol. 17, no. 3, pp. 1153– 1171, 2025

    work page 2025

  39. [40]

    Identifying and predicting heat stress events for grazing dairy cows using rumen temperature boluses. jds com- mun. 5: 431–435,

    S. Woodward, J. Edwards, K. Verhoek, and J. Jago, “Identifying and predicting heat stress events for grazing dairy cows using rumen temperature boluses. jds com- mun. 5: 431–435,” 2024

    work page 2024

  40. [41]

    Optimized machine learning models for predicting core body temperature in dairy cows: enhancing accuracy and interpretability for practical livestock management,

    D. Li, G. Yan, F. Li, H. Lin, H. Jiao, H. Han, and W. Liu, “Optimized machine learning models for predicting core body temperature in dairy cows: enhancing accuracy and interpretability for practical livestock management,” Animals, vol. 14, no. 18, p. 2724, 2024

    work page 2024

  41. [42]

    A deep learning model to forecast cattle heat stress,

    N. H. Chapman, A. Chlingaryan, P. C. Thomson, S. Lo- max, M. A. Islam, A. K. Doughty, and C. E. Clark, “A deep learning model to forecast cattle heat stress,”Com- puters and electronics in agriculture, vol. 211, p. 107932, 2023

    work page 2023

  42. [43]

    Iumenta: A generic framework for ani- mal digital twins within the open digital twin platform,

    A. Youssef, K. Vodorezova, Y. Aarts, W. E. K. Agbeti, A. P. Palstra, E. Foekema, L. Aguilar, R. d. S. Torres, and J. Gr¨ ubel, “Iumenta: A generic framework for ani- mal digital twins within the open digital twin platform,” 2024, arXiv preprint arXiv:2411.10466

    work page arXiv 2024

  43. [44]

    Multimodal behavior recognition for dairy cow digital twin construction under incomplete modali- ties: A modality mapping completion network approach,

    Y. Zhang, Y. Zhang, M. Gao, X. Wang, B. Dai, and W. Shen, “Multimodal behavior recognition for dairy cow digital twin construction under incomplete modali- ties: A modality mapping completion network approach,” Artificial Intelligence in Agriculture, vol. 15, pp. 459–469, 2025

    work page 2025

  44. [45]

    A digital twin of tropical dairy cattle 22 reveals the impact of temperature stress on feeding be- haviour,

    S. Kemp, M. Salavati, R. Dhulipala, A. Paliwal, K. Singh, R. Shin, T. Heo, N. Kim, N. Kipchirchir, I. Gluecks, and W. Sospeter, “A digital twin of tropical dairy cattle 22 reveals the impact of temperature stress on feeding be- haviour,” CGIAR Accelerator on Digital Transformation, Tech. Rep., 2025

    work page 2025

  45. [46]

    Bridging re- search and practice using digital twin modeling to miti- gate methane emissions across dairy production stages,

    C. Castillo, R. Mui˜ no, J. L. Benedito, E. N. Martinez, O. Lopez, G. Salcedo, and J. Hernandez, “Bridging re- search and practice using digital twin modeling to miti- gate methane emissions across dairy production stages,” Journal of Animal Science, vol. 103, p. skaf329, 2025

    work page 2025

  46. [47]

    Insights into thermal stress effects on performance and behavior of grazing cattle via mul- timodal sensor monitoring,

    R. Eckhardt, R. Arablouei, K. McCosker, G. Bishop- Hurley, N. Bagnall, B. Hayes, A. Reverter, A. Ingham, and H. Bernhardt, “Insights into thermal stress effects on performance and behavior of grazing cattle via mul- timodal sensor monitoring,”Scientific Reports, vol. 15, no. 1, p. 27941, 2025

    work page 2025

  47. [48]

    Digital twins in agriculture: challenges and opportunities for environmental sustainability,

    W. Purcell, T. Neubauer, and K. Mallinger, “Digital twins in agriculture: challenges and opportunities for environmental sustainability,”Current Opinion in En- vironmental Sustainability, vol. 61, p. 101252, 2023

    work page 2023

  48. [49]

    From reality to virtuality: Revolutionizing livestock farming through digital twins,

    E. Arulmozhi, N. C. Deb, N. Tamrakar, D. Y. Kang, M. Y. Kang, J. Kook, J. K. Basak, and H. T. Kim, “From reality to virtuality: Revolutionizing livestock farming through digital twins,”Agriculture, vol. 14, no. 12, p. 2231, 2024

    work page 2024

  49. [50]

    Holstein friesian dairy cattle physiological adaptation in balochis- tan: A review,

    M. N. Jan, A. W. Baloch, and W. Irfan, “Holstein friesian dairy cattle physiological adaptation in balochis- tan: A review,”Agricultural Reviews, vol. 46, no. 4, pp. 621–629, 2025

    work page 2025

  50. [51]

    A mechanistic thermal balance model of dairy cattle,

    J. Li, V. Narayanan, E. Kebreab, S. Dikmen, and J. G. Fadel, “A mechanistic thermal balance model of dairy cattle,”Biosystems Engineering, vol. 209, pp. 256–270, 2021

    work page 2021

  51. [52]

    Development and evaluation of a thermoreg- ulatory model for predicting thermal responses of dairy cows,

    M. Zhou, P. W. G. Koerkamp, T. T. Huynh, and A. J. Aarnink, “Development and evaluation of a thermoreg- ulatory model for predicting thermal responses of dairy cows,”Biosystems Engineering, vol. 223, pp. 295–308, 2022

    work page 2022

  52. [53]

    Thermodynamic assess- ment of heat stress in dairy cattle: lessons from human biometeorology,

    S. Foroushani and T. Amon, “Thermodynamic assess- ment of heat stress in dairy cattle: lessons from human biometeorology,”International Journal of Biometeorol- ogy, vol. 66, no. 9, pp. 1811–1827, 2022

    work page 2022

  53. [54]

    A compre- hensive review of digital twins technology in agriculture,

    R. Zhang, H. Zhu, Q. Chang, and Q. Mao, “A compre- hensive review of digital twins technology in agriculture,” Agriculture, vol. 15, no. 9, p. 903, 2025

    work page 2025

  54. [55]

    A systematic review of hidden markov models and their applica- tions

    B. Mor, S. Garhwal, and A. Kumar, “A systematic review of hidden markov models and their applica- tions.”Archives of computational methods in engineer- ing, vol. 28, no. 3, 2021

    work page 2021

  55. [56]

    Sensor fusion,

    J. Z. Sasiadek, “Sensor fusion,”Annual Reviews in Con- trol, vol. 26, no. 2, pp. 203–228, 2002

    work page 2002

  56. [57]

    Gaussian processes for machine learning,

    M. Seeger, “Gaussian processes for machine learning,” International journal of neural systems, vol. 14, no. 02, pp. 69–106, 2004

    work page 2004

  57. [58]

    Computational architec- tures for precision dairy nutrition digital twins: A tech- nical review and implementation framework,

    S. Rao and S. Neethirajan, “Computational architec- tures for precision dairy nutrition digital twins: A tech- nical review and implementation framework,”Sensors, vol. 25, no. 16, p. 4899, 2025

    work page 2025

  58. [59]

    Physiological responses of cattle to heat stress,

    U. Farooq, H. A. Samad, F. Shehzad, and A. Qayyum, “Physiological responses of cattle to heat stress,”World Applied Sciences Journal, vol. 8, pp. 38–43, 2010

    work page 2010

  59. [60]

    Lightgbm: A highly efficient gradient boosting decision tree,

    G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T.-Y. Liu, “Lightgbm: A highly efficient gradient boosting decision tree,” inAdvances in Neu- ral Information Processing Systems (NeurIPS), vol. 30, 2017

    work page 2017

  60. [61]

    Optuna: A next-generation hyperparameter optimiza- tion framework,

    T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Moyama, “Optuna: A next-generation hyperparameter optimiza- tion framework,” inProceedings of the 25th ACM SIGKDD International Conference on Knowledge Dis- covery and Data Mining (KDD), 2019, pp. 2623–2631

    work page 2019

  61. [62]

    Bootstrap resampling,

    P. M. Dixon, “Bootstrap resampling,” inEncyclopedia of Environmetrics, 2006

    work page 2006

  62. [63]

    Mmcows: A multi- modal dataset for dairy cattle monitoring,

    H. Vu, O. C. Prabhune, U. Raskar, D. Panditharatne, H. Chung, C. Choi, and Y. Kim, “Mmcows: A multi- modal dataset for dairy cattle monitoring,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 37, 2024, pp. 59 451–59 467

    work page 2024

  63. [64]

    Heat stress in dairy cows: Impacts, identification, and mitiga- tion strategies—a review,

    C. P. Oliveira, F. C. de Sousa, A. L. da Silva, ´E. B. Schultz, R. I. V. Londo˜ no, and P. A. R. de Souza, “Heat stress in dairy cows: Impacts, identification, and mitiga- tion strategies—a review,”Animals, vol. 15, no. 2, 2025

    work page 2025

  64. [65]

    An in vivo model to assess the thermoregulatory response of lactating holsteins to an acute heat stress event occurring after a pharmacologically-induced lh surge,

    C. R. Abbott, A. M. Saxton, L. A. Rispoli, R. R. Payton, K. G. Pohler, F. N. Schrick, and J. L. Edwards, “An in vivo model to assess the thermoregulatory response of lactating holsteins to an acute heat stress event occurring after a pharmacologically-induced lh surge,”Journal of Thermal Biology, vol. 78, pp. 247–256, 2018

    work page 2018

  65. [66]

    Development and evaluation of equations for prediction of feed intake for lactating holstein dairy cows,

    D. K. Roseler, D. G. Fox, L. E. Chase, A. N. Pell, and W. C. Stone, “Development and evaluation of equations for prediction of feed intake for lactating holstein dairy cows,”Journal of Dairy Science, vol. 80, no. 5, pp. 878– 893, 1997

    work page 1997

  66. [67]

    Assessing the impact of thermoregulatory mineral sup- plementation on thermal comfort in lactating holstein cows,

    R. Felini, D. Cavallini, G. Buonaiuto, and T. Bordin, “Assessing the impact of thermoregulatory mineral sup- plementation on thermal comfort in lactating holstein cows,”Veterinary and Animal Science, vol. 24, 2024

    work page 2024

  67. [68]

    Why do tree-based models still outperform deep learning on typ- ical tabular data?

    L. Grinsztajn, E. Oyallon, and G. Varoquaux, “Why do tree-based models still outperform deep learning on typ- ical tabular data?” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022, pp. 507– 520

    work page 2022

  68. [69]

    Tabular data: Deep learning is not all you need,

    R. Shwartz-Ziv and A. Armon, “Tabular data: Deep learning is not all you need,”Information Fusion, vol. 81, 23 pp. 84–90, 2022. 24

    work page 2022