pith. sign in

arxiv: 2606.20329 · v1 · pith:X366TZM7new · submitted 2026-06-18 · 💻 cs.LG · physics.geo-ph

Constrained hybrid modelling to predict microbial dynamics and organic matter turnover in soil systems

Paul Collart , Juergen Gall , Andrea Schnepf , Holger Pagel , Lars Doorenbos This is my paper

Pith reviewed 2026-06-26 17:42 UTC · model grok-4.3

classification 💻 cs.LG physics.geo-ph
keywords hybrid modelingsoil organic mattermicrobial dynamicsmetagenomicsneural networksprocess-based modelsecological constraintsbiokinetic parameters
0
0 comments X

The pith

A neural network maps metagenomic traits to biokinetic parameters in a soil organic matter model while ecological constraints keep unmeasurable states realistic.

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

The paper establishes a hybrid framework that feeds metagenome-inferred functional traits into a neural network to set the biokinetic parameters of a process-based soil model. Ecological constraints drawn from theory and literature are imposed so that even variables without direct observations behave plausibly. This matters because microbial activity largely controls soil carbon turnover, yet conventional models lack enough data to set their parameters accurately. Tests on synthetic trait datasets of varying complexity and on real sequencing data show gains over baselines, including when training data are scarce.

Core claim

The central claim is that a neural network can derive realistic biokinetic parameter values for a process-based soil organic matter turnover model from metagenome-inferred functional traits, with ecological constraints ensuring plausible behavior of non-observed state variables, and that this yields better performance than baselines even on small training datasets.

What carries the argument

The constrained hybrid modeling framework in which a neural network predicts biokinetic parameters from genomic trait data and ecological constraints regularize the mapping to keep non-observed dynamics realistic.

If this is right

  • Improved accuracy for unmeasurable state variables in the process-based model.
  • Outperformance over multiple baselines on both synthetic and real genomic trait data.
  • Effective learning of dynamics even when training datasets are small.
  • Realistic outputs maintained by the ecological constraints for all state variables.

Where Pith is reading between the lines

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

  • The same mapping could be tested on additional omics layers such as metatranscriptomics to check whether trait-to-parameter translation improves further.
  • Field-scale application might allow the model to forecast how different land-management choices alter carbon storage trajectories.
  • Cross-validation across soil types from contrasting climates would show how far the learned trait-to-parameter relationships transfer beyond the original training conditions.

Load-bearing premise

A neural network can learn a generalizable mapping from metagenome-inferred functional traits to biokinetic parameters of the process-based model when regularized by ecological constraints.

What would settle it

Predictions that deviate substantially from measured microbial dynamics or organic matter turnover rates on independent real soil datasets withheld from training.

Figures

Figures reproduced from arXiv: 2606.20329 by Andrea Schnepf, Holger Pagel, Juergen Gall, Lars Doorenbos, Paul Collart.

Figure 1
Figure 1. Figure 1: The HySoMi hybrid modelling framework for soil carbon cycling predictions. The goal is to learn a mapping gw from genomic data, aggregated into a vector of genomic traits [T1, T2, ..., Ti], to biokinetic parameters θbio. The biokinetic parameters, however, cannot be measured. Instead, time series of CO2 are the only available measurements. To link the unknown biokinetic parameters θbio to measurable variab… view at source ↗
Figure 2
Figure 2. Figure 2: MSE for each state variable on test data. Both HySoMi and Unconstrained only use CO2 during training. The All states scenario is trained with every state variable. State vari￾ables are described in Table A.3. Despite only being trained with CO2, HySoMi achieves comparable performance to the unrealis￾tic All states model. hybrid model retains interpretable intermediate variables, which are essential for pro… view at source ↗
Figure 6
Figure 6. Figure 6: shows that HySoMi predicts realistic behavior even when MSEhidden for a particular sample is higher than average. Compared to the Unconstrained scenario, the model shows realistic behavior for inactive bacteria (Bi) in particular. Even when not fitting directly to this state variable, we observe a sharp transition to dormancy and a relatively stable dormant population afterwards, which is expected for the … view at source ↗
Figure 5
Figure 5. Figure 5: Constraint satisfaction on validation set after training. Only a single constraint stays active after training for HySoMi. The Unconstrained model is not informed by its training data on the P6 and C3 constraints, leading to unrealistic model behavior. Constraints are described in [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

Soil microorganisms control organic matter cycling and largely determine how soil systems can cope with and mitigate climate change and environmental threats. Representing microbial dynamics in process-based soil models is therefore critical to predict carbon cycling in soils, albeit highly challenging to inform from data. One promising approach to improve their parametrisation is the integration of genomic data, yet modelling the complex and unknown relationship between genomes and the processes the microbes are driving is an unsolved problem. In this work, we present the first hybrid modeling framework for deriving biokinetic parameter values of a process-based soil organic matter turnover model from metagenome-inferred functional traits based on DNA sequencing data. Our model predicts biokinetic parameters of the process-based model from genomic trait data with a neural network and integrates constraints from ecological theory and literature to ensure realistic behavior, even of non-observed state variables. We evaluate our method on synthetic genomic trait datasets of varying complexity and on real data, showing that our approach improves performance over multiple baselines and learns the dynamics of unmeasurable components of the process-based model effectively, even for small training datasets.

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 paper presents the first hybrid modeling framework that uses a neural network to predict biokinetic parameters of a process-based soil organic matter turnover model from metagenome-inferred functional traits derived from DNA sequencing data. Ecological constraints from theory and literature are incorporated to regularize behavior of unobserved state variables. Evaluation is performed on synthetic genomic trait datasets of varying complexity and on real soil data, with claims of improved performance over multiple baselines and effective learning of unmeasurable component dynamics even for small training datasets.

Significance. If the central mapping generalizes, the approach could meaningfully advance data-informed parametrization of microbial dynamics in soil carbon models, addressing a key challenge in predicting responses to climate and environmental change. The explicit use of ecological constraints to regularize non-observed states is a constructive strength that distinguishes the method from purely data-driven alternatives.

major comments (2)
  1. [Real data evaluation] Real-data evaluation (as described in the abstract and methods): the reported improvements on real soil datasets lack an independent hold-out set drawn from a different soil type, location, or environmental regime. Without such a test or proxy measurements for the hidden states, it is not possible to confirm that the NN mapping from functional traits to biokinetic parameters transfers outside the training distribution rather than overfitting to the observed variables under the ecological regularizers.
  2. [Synthetic experiments] Synthetic-to-real transfer (evaluation sections): while synthetic experiments can demonstrate recovery by construction, the manuscript does not report quantitative metrics (e.g., parameter recovery error or trajectory error on held-out synthetic regimes) that would establish the conditions under which the constrained NN mapping remains accurate when the underlying trait-to-parameter relationship deviates from the training distribution.
minor comments (2)
  1. [Abstract] The abstract lists 'multiple baselines' without naming them or indicating whether they include both purely process-based and unconstrained neural hybrids; this should be clarified for reproducibility.
  2. [Methods] Notation for the ecological constraints and the precise form of the regularization term in the loss function should be defined explicitly (e.g., as an equation) rather than described at a high level.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We appreciate the referee's detailed and constructive feedback on our manuscript. We address the major comments below and outline the revisions we plan to make.

read point-by-point responses

  1. Referee: [Real data evaluation] Real-data evaluation (as described in the abstract and methods): the reported improvements on real soil datasets lack an independent hold-out set drawn from a different soil type, location, or environmental regime. Without such a test or proxy measurements for the hidden states, it is not possible to confirm that the NN mapping from functional traits to biokinetic parameters transfers outside the training distribution rather than overfitting to the observed variables under the ecological regularizers.

    Authors: We agree that an independent hold-out set from a different soil type or environmental regime would strengthen the evidence for transferability of the NN mapping. Our evaluation on real data relies on cross-validation within the available dataset, which demonstrates improved performance over baselines but does not fully address out-of-distribution generalization. We will revise the manuscript to include a more explicit discussion of this limitation and explore the possibility of incorporating additional real-world datasets for validation if feasible. revision: partial

  2. Referee: [Synthetic experiments] Synthetic-to-real transfer (evaluation sections): while synthetic experiments can demonstrate recovery by construction, the manuscript does not report quantitative metrics (e.g., parameter recovery error or trajectory error on held-out synthetic regimes) that would establish the conditions under which the constrained NN mapping remains accurate when the underlying trait-to-parameter relationship deviates from the training distribution.

    Authors: We acknowledge the value of reporting additional quantitative metrics on held-out synthetic regimes to assess performance under deviations from the training distribution. We will update the evaluation section to include parameter recovery errors and trajectory errors for such cases, thereby better characterizing the robustness of the constrained hybrid model. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical claims rest on independent evaluation

full rationale

The paper describes a hybrid framework in which a neural network maps metagenomic traits to biokinetic parameters of a process-based model, with regularization drawn from external ecological theory and literature. Performance is assessed by direct comparison to baselines on both synthetic and real datasets, including the ability to track unmeasurable states. No equations, parameter-fitting steps, or citations are presented that reduce the central mapping or predictions to the training data by construction. The load-bearing elements (NN architecture, ecological constraints, generalization) are stated as external inputs rather than self-derived, making the reported improvements falsifiable against held-out data and independent benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no concrete equations or sections from which to extract free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5729 in / 945 out tokens · 23422 ms · 2026-06-26T17:42:09.081382+00:00 · methodology

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

64 extracted references · 42 canonical work pages · 2 internal anchors

  1. [1]

    Nature Climate Change , author =

    Global soil carbon projections are improved by modelling microbial processes , volume =. Nature Climate Change , author =. 2013 , note =. doi:10.1038/nclimate1951 , abstract =

    work page doi:10.1038/nclimate1951 2013

  2. [2]

    Biogeosciences , author =

    A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations , volume =. Biogeosciences , author =. 2023 , note =. doi:10.5194/bg-20-2671-2023 , abstract =

    work page doi:10.5194/bg-20-2671-2023 2023

  3. [3]

    Nature Communications , author =

    From calibration to parameter learning:. Nature Communications , author =. 2021 , note =. doi:10.1038/s41467-021-26107-z , abstract =

    work page doi:10.1038/s41467-021-26107-z 2021

  4. [4]

    Nature Microbiology , author =

    Microbial ecology:. Nature Microbiology , author =. 2016 , note =. doi:10.1038/nmicrobiol.2015.28 , abstract =

    work page doi:10.1038/nmicrobiol.2015.28 2016

  5. [5]

    Deep Learning - whole book,

    Deep learning , volume =. Nature , author =. 2015 , note =. doi:10.1038/nature14539 , abstract =

    work page doi:10.1038/nature14539 2015

  6. [6]

    A Differentiable Programming System to Bridge Machine Learning and Scientific Computing

    Innes, Mike and Edelman, Alan and Fischer, Keno and Rackauckas, Chris and Saba, Elliot and Shah, Viral B. and Tebbutt, Will , month = jul, year =. A. doi:10.48550/arXiv.1907.07587 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1907.07587 1907

  7. [7]

    NatureReviewsEarth&Environment 4, 552–567

    Differentiable modelling to unify machine learning and physical models for geosciences , volume =. Nature Reviews Earth & Environment , author =. 2023 , note =. doi:10.1038/s43017-023-00450-9 , abstract =

    work page doi:10.1038/s43017-023-00450-9 2023

  8. [8]

    Advances in

    Paszke, Adam and Gross, Sam and Massa, Francisco and Lerer, Adam and Bradbury, James and Chanan, Gregory and Killeen, Trevor and Lin, Zeming and Gimelshein, Natalia and Antiga, Luca and Desmaison, Alban and Kopf, Andreas and Yang, Edward and DeVito, Zachary and Raison, Martin and Tejani, Alykhan and Chilamkurthy, Sasank and Steiner, Benoit and Fang, Lu an...

  9. [9]

    He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian , month = jun, year =. Deep. 2016. doi:10.1109/CVPR.2016.90 , abstract =

    work page doi:10.1109/cvpr.2016.90 2016

  10. [10]

    Approximate

    Narasimhan, Harikrishna and Cotter, Andrew and Zhou, Yichen and Wang, Serena and Guo, Wenshuo , year =. Approximate. Advances in

  11. [12]

    IEEE Potentials , author =

    Feed-forward neural networks , volume =. IEEE Potentials , author =. 1994 , keywords =. doi:10.1109/45.329294 , abstract =

    work page doi:10.1109/45.329294 1994

  12. [13]

    Agarap, Abien Fred , month = feb, year =. Deep. doi:10.48550/arXiv.1803.08375 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1803.08375

  13. [14]

    Raissi, P

    Physics-informed neural networks:. Journal of Computational Physics , author =. 2019 , keywords =. doi:10.1016/j.jcp.2018.10.045 , abstract =

    work page doi:10.1016/j.jcp.2018.10.045 2019

  14. [15]

    Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang

    Physics-informed machine learning , volume =. Nature Reviews Physics , author =. 2021 , note =. doi:10.1038/s42254-021-00314-5 , abstract =

    work page doi:10.1038/s42254-021-00314-5 2021

  15. [18]

    He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian , year =. Delving. International Conference on Computer Vision , url =

  16. [19]

    International Conference on Machine Learning , pages =

    Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , author =. International Conference on Machine Learning , pages =. 2015 , editor =

    2015

  17. [20]

    and Jung, M

    Kraft, B. and Jung, M. and K\"orner, M. and Koirala, S. and Reichstein, M. , TITLE =. Hydrology and Earth System Sciences , VOLUME =. 2022 , NUMBER =

    2022

  18. [21]

    Water Resources Research , volume =

    Schmidt, Lennart and Heße, Falk and Attinger, Sabine and Kumar, Rohini , title =. Water Resources Research , volume =. doi:https://doi.org/10.1029/2019WR025924 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2019WR025924 , note =

    work page doi:10.1029/2019wr025924

  19. [22]

    Environmental Research Letters , abstract =

    ElGhawi, Reda and Kraft, Basil and Reimers, Christian and Reichstein, Markus and Körner, Marco and Gentine, Pierre and Winkler, Alexander J , title =. Environmental Research Letters , abstract =. 2023 , month =. doi:10.1088/1748-9326/acbbe0 , url =

    work page doi:10.1088/1748-9326/acbbe0 2023

  20. [23]

    Chen, Ricky T. Q. , title=. 2018 , url=

    2018

  21. [24]

    2017 , eprint=

    Adam: A Method for Stochastic Optimization , author=. 2017 , eprint=

    2017

  22. [25]

    Soil carbon and nitrogen mineralization: Theory and models across scales , journal =

    Stefano Manzoni and Amilcare Porporato , keywords =. Soil carbon and nitrogen mineralization: Theory and models across scales , journal =. 2009 , issn =. doi:https://doi.org/10.1016/j.soilbio.2009.02.031 , url =

    work page doi:10.1016/j.soilbio.2009.02.031 2009

  23. [26]

    Ecological Modelling , author =

    Incorporating dormancy in dynamic microbial community models , volume =. Ecological Modelling , author =. 2011 , keywords =. doi:10.1016/j.ecolmodel.2011.07.006 , abstract =

    work page doi:10.1016/j.ecolmodel.2011.07.006 2011

  24. [27]

    Frontiers in Environmental Science , author =

    Spatial. Frontiers in Environmental Science , author =. 2020 , file =

    2020

  25. [28]

    Wieder, W. R. and Grandy, A. S. and Kallenbach, C. M. and Taylor, P. G. and Bonan, G. B. , TITLE =. Geoscientific Model Development , VOLUME =. 2015 , NUMBER =

    2015

  26. [29]

    Proceedings of the Asian Conference on Computer Vision (ACCV) , month =

    Sargeant, James and Teng, Shyh Wei and Murshed, Manzur and Paul, Manoranjan and Brennan, David , title =. Proceedings of the Asian Conference on Computer Vision (ACCV) , month =. 2024 , pages =

    2024

  27. [30]

    Acta Mechanica Sinica , author =

    Physics-informed neural networks (. Acta Mechanica Sinica , author =. 2021 , pages =. doi:10.1007/s10409-021-01148-1 , abstract =

    work page doi:10.1007/s10409-021-01148-1 2021

  28. [31]

    Journal of Geophysical Research: Biogeosciences , volume =

    Chandel, Aneesh Kumar and Jiang, Lifen and Luo, Yiqi , title =. Journal of Geophysical Research: Biogeosciences , volume =. doi:https://doi.org/10.1029/2023JG007436 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2023JG007436 , note =

    work page doi:10.1029/2023jg007436

  29. [32]

    Nature Communications , author =

    Gene-informed decomposition model predicts lower soil carbon loss due to persistent microbial adaptation to warming , volume =. Nature Communications , author =. 2020 , note =. doi:10.1038/s41467-020-18706-z , abstract =

    work page doi:10.1038/s41467-020-18706-z 2020

  30. [33]

    Journal of Geophysical Research: Biogeosciences , volume =

    Aboelyazeed, Doaa and Xu, Chonggang and Gu, Lianhong and Luo, Xiangzhong and Liu, Jiangtao and Lawson, Kathryn and Shen, Chaopeng , title =. Journal of Geophysical Research: Biogeosciences , volume =. doi:https://doi.org/10.1029/2024JG008552 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2024JG008552 , note =

    work page doi:10.1029/2024jg008552

  31. [34]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Kendall, Alex and Gal, Yarin and Cipolla, Roberto , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  32. [35]

    Coupling of microbial-explicit model and machine learning improves the prediction and turnover process simulation of soil organic carbon , journal =

    Xuebin Xu and Xianting Wang and Ping Zhou and Zhenke Zhu and Liang Wei and Shuang Wang and Periyasamy Rathinapriya and Qicheng Bei and Jinfei Feng and Fuping Fang and Jianping Chen and Tida Ge , keywords =. Coupling of microbial-explicit model and machine learning improves the prediction and turnover process simulation of soil organic carbon , journal =. ...

    work page doi:10.1016/j.csag.2024.100001 2024

  33. [36]

    and Pagel, Holger and Kügler, Philipp and Streck, Thilo , title =

    Marschmann, Gianna L. and Pagel, Holger and Kügler, Philipp and Streck, Thilo , title =. 2019 , journal =. doi:10.1016/j.envsoft.2019.104518 , url =

    work page doi:10.1016/j.envsoft.2019.104518 2019

  34. [37]

    , TITLE=

    Karaoz, Ulas and Brodie, Eoin L. , TITLE=. Frontiers in Bioinformatics , VOLUME=. 2022 , URL=. doi:10.3389/fbinf.2022.918853 , ISSN=

    work page doi:10.3389/fbinf.2022.918853 2022

  35. [38]

    ISME Communications , volume =

    Dragone, Nicholas B and Hoffert, Michael and Strickland, Michael S and Fierer, Noah , title =. ISME Communications , volume =. 2024 , month =. doi:10.1093/ismeco/ycae081 , url =

    work page doi:10.1093/ismeco/ycae081 2024

  36. [39]

    Shapiro and Charles S

    Humberto Blanco-Canqui and Charles A. Shapiro and Charles S. Wortmann and Rhae A. Drijber and Martha Mamo and Tim M. Shaver and Richard B. Ferguson , title =. Journal of Soil and Water Conservation , volume =. 2013 , publisher =. doi:10.2489/jswc.68.5.129A , URL =

    work page doi:10.2489/jswc.68.5.129a 2013

  37. [40]

    Plant and Soil , author =

    The variation of soil microbial respiration with depth in relation to soil carbon composition , volume =. Plant and Soil , author =. 2005 , keywords =. doi:10.1007/s11104-004-0278-4 , abstract =

    work page doi:10.1007/s11104-004-0278-4 2005

  38. [41]

    Spatial substrate heterogeneity limits microbial growth as revealed by the joint experimental quantification and modeling of carbon and heat fluxes , journal =

    Martin-Georg Endress and Fatemeh Dehghani and Sergey Blagodatsky and Thomas Reitz and Steffen Schlüter and Evgenia Blagodatskaya , keywords =. Spatial substrate heterogeneity limits microbial growth as revealed by the joint experimental quantification and modeling of carbon and heat fluxes , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.soilbio....

    work page doi:10.1016/j.soilbio.2024.109509 2024

  39. [42]

    and Hobbie, Erik A

    Hobbie, John E. and Hobbie, Erik A. , TITLE=. Frontiers in Microbiology , VOLUME=. 2013 , URL=. doi:10.3389/fmicb.2013.00324 , ISSN=

    work page doi:10.3389/fmicb.2013.00324 2013

  40. [43]

    The effects of glucose loading rates on bacterial and fungal growth in soil , journal =

    Stephanie Reischke and Johannes Rousk and Erland Bååth , keywords =. The effects of glucose loading rates on bacterial and fungal growth in soil , journal =. 2014 , issn =. doi:https://doi.org/10.1016/j.soilbio.2013.12.011 , url =

    work page doi:10.1016/j.soilbio.2013.12.011 2014

  41. [44]

    Biology Bulletin Reviews , author =

    Metabarcoding and. Biology Bulletin Reviews , author =. 2021 , pages =. doi:10.1134/S2079086421010084 , abstract =

    work page doi:10.1134/s2079086421010084 2021

  42. [45]

    , title =

    Manzoni, Stefano and Taylor, Philip and Richter, Andreas and Porporato, Amilcare and Ågren, Göran I. , title =. New Phytologist , volume =. doi:https://doi.org/10.1111/j.1469-8137.2012.04225.x , url =. https://nph.onlinelibrary.wiley.com/doi/pdf/10.1111/j.1469-8137.2012.04225.x , abstract =

    work page doi:10.1111/j.1469-8137.2012.04225.x 2012

  43. [46]

    Modeling coupled pesticide degradation and organic matter turnover: From gene abundance to process rates , journal =

    Holger Pagel and Christian Poll and Joachim Ingwersen and Ellen Kandeler and Thilo Streck , keywords =. Modeling coupled pesticide degradation and organic matter turnover: From gene abundance to process rates , journal =. 2016 , issn =. doi:https://doi.org/10.1016/j.soilbio.2016.09.014 , url =

    work page doi:10.1016/j.soilbio.2016.09.014 2016

  44. [47]

    Trait-based modeling of microbial interactions and carbon turnover in the rhizosphere , journal =

    Ahmet Kürşad Sırcan and Thilo Streck and Andrea Schnepf and Mona Giraud and Adrian Lattacher and Ellen Kandeler and Christian Poll and Holger Pagel , keywords =. Trait-based modeling of microbial interactions and carbon turnover in the rhizosphere , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.soilbio.2024.109698 , url =

    work page doi:10.1016/j.soilbio.2024.109698 2025

  45. [48]

    FEMS Microbiology Ecology , volume =

    Stolpovsky, Konstantin and Fetzer, Ingo and Van Cappellen, Philippe and Thullner, Martin , title =. FEMS Microbiology Ecology , volume =. 2016 , month =. doi:10.1093/femsec/fiw071 , url =

    work page doi:10.1093/femsec/fiw071 2016

  46. [49]

    Modeling ecosystem-scale carbon dynamics in soil: The microbial dimension , journal =

    Joshua Schimel , keywords =. Modeling ecosystem-scale carbon dynamics in soil: The microbial dimension , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.soilbio.2023.108948 , url =

    work page doi:10.1016/j.soilbio.2023.108948 2023

  47. [50]

    and Schnepf, A

    Vereecken, H. and Schnepf, A. and Hopmans, J.W. and Javaux, M. and Or, D. and Roose, T. and Vanderborght, J. and Young, M.H. and Amelung, W. and Aitkenhead, M. and Allison, S.D. and Assouline, S. and Baveye, P. and Berli, M. and Brüggemann, N. and Finke, P. and Flury, M. and Gaiser, T. and Govers, G. and Ghezzehei, T. and Hallett, P. and Hendricks Fransse...

    work page doi:10.2136/vzj2015.09.0131

  48. [51]

    Nature Microbiology , author =

    Predictions of rhizosphere microbiome dynamics with a genome-informed and trait-based energy budget model , volume =. Nature Microbiology , author =. 2024 , note =. doi:10.1038/s41564-023-01582-w , abstract =

    work page doi:10.1038/s41564-023-01582-w 2024

  49. [52]

    Kothawala, D. N. and Moore, T. R. and Hendershot, W. H. , title =. Soil Science Society of America Journal , volume =. doi:https://doi.org/10.2136/sssaj2008.0254 , url =. https://acsess.onlinelibrary.wiley.com/doi/pdf/10.2136/sssaj2008.0254 , abstract =

    work page doi:10.2136/sssaj2008.0254

  50. [53]

    2018 , eprint=

    Auxiliary Tasks in Multi-task Learning , author=. 2018 , eprint=

    2018

  51. [54]

    and Wieder, William R

    Bradford, Mark A. and Wieder, William R. and Bonan, Gordon B. and Fierer, Noah and Raymond, Peter A. and Crowther, Thomas W. , month = aug, year =. Managing uncertainty in soil carbon feedbacks to climate change , volume =. Nature Climate Change , publisher =. doi:10.1038/nclimate3071 , abstract =

    work page doi:10.1038/nclimate3071

  52. [55]

    2018 , eprint=

    Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics , author=. 2018 , eprint=

    2018

  53. [56]

    and Ballabio, C

    Orgiazzi, A. and Ballabio, C. and Panagos, P. and Jones, A. and Fernández-Ugalde, O. , title =. European Journal of Soil Science , volume =. doi:https://doi.org/10.1111/ejss.12499 , url =. https://bsssjournals.onlinelibrary.wiley.com/doi/pdf/10.1111/ejss.12499 , abstract =

    work page doi:10.1111/ejss.12499

  54. [57]

    Sulman and Jeffrey S

    Alejandro Salazar and Benjamin N. Sulman and Jeffrey S. Dukes , keywords =. Microbial dormancy promotes microbial biomass and respiration across pulses of drying-wetting stress , journal =. 2018 , issn =. doi:https://doi.org/10.1016/j.soilbio.2017.10.017 , url =

    work page doi:10.1016/j.soilbio.2017.10.017 2018

  55. [58]

    International Conference on Learning Representations , year=

    Adam: A method for stochastic optimization , author=. International Conference on Learning Representations , year=

  56. [59]

    and Dacal, Marina and Hartley, Iain P

    García-Palacios, Pablo and Crowther, Thomas W. and Dacal, Marina and Hartley, Iain P. and Reinsch, Sabine and Rinnan, Riikka and Rousk, Johannes and van den Hoogen, Johan and Ye, Jian-Sheng and Bradford, Mark A. , month = jul, year =. Evidence for large microbial-mediated losses of soil carbon under anthropogenic warming , volume =. Nature Reviews Earth &...

    work page doi:10.1038/s43017-021-00178-4

  57. [60]

    Langley , title =

    P. Langley , title =. Proceedings of the 17th International Conference on Machine Learning (ICML 2000) , address =. 2000 , pages =

    2000

  58. [61]

    T. M. Mitchell. The Need for Biases in Learning Generalizations. 1980

    1980

  59. [62]

    M. J. Kearns , title =

  60. [63]

    Machine Learning: An Artificial Intelligence Approach, Vol. I. 1983

    1983

  61. [64]

    R. O. Duda and P. E. Hart and D. G. Stork. Pattern Classification. 2000

    2000

  62. [65]

    Suppressed for Anonymity , author=

  63. [66]

    Newell and P

    A. Newell and P. S. Rosenbloom. Mechanisms of Skill Acquisition and the Law of Practice. Cognitive Skills and Their Acquisition. 1981

    1981

  64. [67]

    A. L. Samuel. Some Studies in Machine Learning Using the Game of Checkers. IBM Journal of Research and Development. 1959

    1959