pith. sign in

arxiv: 2606.21179 · v1 · pith:NTX4MKJSnew · submitted 2026-06-19 · 💻 cs.LG

Rejections Based on Predictive Uncertainty Enable Reliable Routine Soil Spectroscopy

Jonas Schmidinger , Robin Gebbers , Marc-Olivier Gasser , Viacheslav Barkov , G.Mick Wu , Viacheslav I. Adamchuk This is my paper

Pith reviewed 2026-06-26 14:29 UTC · model grok-4.3

classification 💻 cs.LG
keywords soil spectroscopypredictive uncertaintyreject optionfoundation modelsTabPFNmachine learningagricultural measurementsvisible-near-infrared
0
0 comments X

The pith

Rejecting high-uncertainty spectroscopic predictions for conventional remeasurement lets labs meet accuracy targets at lower cost.

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

The paper introduces a reject-to-remeasure workflow that first runs visible-near-infrared spectroscopy on soil samples and then uses probabilistic foundation models to flag predictions whose uncertainty exceeds user-set quality limits. Those flagged samples are sent for full laboratory analysis instead. On a Québec regional soil library the method is shown to satisfy chosen accuracy thresholds for multiple soil properties while lowering the fraction of samples that require the expensive conventional route. A sympathetic reader would care because this directly addresses the main barrier to adopting fast, cheap spectroscopy in everyday agricultural and environmental testing.

Core claim

Reject-to-remeasure pairs probabilistic predictions from TabPFNv2.5 and TabICLv2 with uncertainty thresholds so that only low-uncertainty spectroscopic outputs are accepted; rejected samples are remeasured by conventional laboratory methods. This combination is shown to satisfy user-defined accuracy constraints on a regional Québec soil library while reducing overall measurement costs compared with full conventional analysis of every sample.

What carries the argument

The reject-to-remeasure framework, which uses predictive uncertainty from foundation models to decide whether to accept a spectroscopic prediction or route the sample to conventional remeasurement.

If this is right

  • Choosing higher uncertainty thresholds increases the fraction of samples accepted from spectroscopy while still respecting the accuracy bound.
  • Modern foundation models such as TabPFNv2.5 and TabICLv2 produce uncertainty estimates that support this selective acceptance at scale.
  • Routine laboratory workflows can therefore incorporate spectroscopy for the majority of samples and reserve conventional analysis for the uncertain minority.
  • Overall per-sample measurement cost drops in proportion to the fraction of samples whose uncertainty stays below the chosen threshold.

Where Pith is reading between the lines

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

  • The same rejection logic could be tested on other sensor-based agricultural measurements where ground-truth costs are high.
  • Performance will likely degrade if the spectral library used for training differs markedly in geography or soil type from the samples being tested.
  • Combining the uncertainty signal with additional metadata such as sampling location might further reduce the number of remeasurements needed.

Load-bearing premise

The uncertainty scores produced by the models are well enough calibrated that a chosen rejection threshold actually removes the samples whose predictions would otherwise exceed the accuracy limit.

What would settle it

Run the conventional laboratory method on every sample in a held-out set, then check whether the fraction of accepted predictions that violate the accuracy target matches the rate expected from the uncertainty calibration curves.

read the original abstract

Soil properties relevant to agricultural and environmental applications are conventionally measured using elaborate laboratory methods involving physical and chemical processing. While highly accurate, these conventional methods are costly and time-consuming. In contrast, optical spectroscopy paired with machine learning enables rapid and cost-effective predictions of multiple soil properties. However, spectroscopic modelling is often considered unreliable, as the predictive accuracy varies between soil properties and individual samples. To balance this trade-off between cost and reliability, we introduce reject-to-remeasure: an AI-based measurement framework that combines probabilistic modelling with uncertainty-guided rejection. In this framework, soil samples are first analysed using spectroscopy, after which predictions are rejected if their predictive uncertainty exceeds predefined quality constraints. Rejected samples are subsequently remeasured using conventional laboratory procedures. On a regional visible-near-infrared spectral soil library from Qu\'ebec, we demonstrate that reject-to-remeasure with modern foundation models (TabPFNv2.5 and TabICLv2) can facilitate the integration of optical spectroscopy into routine laboratory workflows while meeting user-defined accuracy requirements and reducing measurement costs.

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 introduces a 'reject-to-remeasure' framework that applies probabilistic foundation models (TabPFNv2.5 and TabICLv2) to visible-near-infrared soil spectra from a Québec regional library. Spectroscopic predictions are accepted only if their predictive uncertainty falls below user-specified thresholds; otherwise the sample is remeasured by conventional laboratory methods. The central empirical claim is that this procedure meets user-defined accuracy targets for multiple soil properties while lowering overall measurement costs relative to full laboratory analysis.

Significance. If the reported uncertainty calibration holds, the framework supplies a practical, controllable mechanism for inserting spectroscopy into routine soil-testing workflows without sacrificing required accuracy. The use of modern tabular foundation models and the explicit cost-accuracy framing are strengths that could accelerate adoption in agricultural and environmental laboratories.

major comments (2)
  1. [Section 3.3 and Figure 4] Section 3.3 and Figure 4: the central claim that rejection at user-chosen uncertainty thresholds reliably removes samples whose predictions would violate accuracy constraints is load-bearing. The manuscript must report, for each soil property, the observed error rate on the accepted subset versus the rejected subset (or an equivalent calibration diagnostic such as reliability diagrams or ECE) so that readers can verify the assumption that uncertainty is sufficiently well-calibrated.
  2. [Section 4.1, Table 2] Section 4.1, Table 2: the cost-reduction percentages are computed relative to a baseline of 100 % laboratory measurements. The paper should state the exact per-sample laboratory cost model and the fraction of samples rejected at each operating point; without these numbers the claimed savings cannot be reproduced or compared across laboratories.
minor comments (2)
  1. [Abstract and Section 2] The abstract and Section 2 refer to 'modern foundation models (TabPFNv2.5 and TabICLv2)' without a brief citation or version note; adding the original references would improve traceability.
  2. [Figure 3] Figure 3 caption should explicitly define the axes of the uncertainty-vs-error scatter plots and state whether the plotted uncertainty is the predictive standard deviation or another quantity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment and constructive comments. We address each major point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Section 3.3 and Figure 4] Section 3.3 and Figure 4: the central claim that rejection at user-chosen uncertainty thresholds reliably removes samples whose predictions would violate accuracy constraints is load-bearing. The manuscript must report, for each soil property, the observed error rate on the accepted subset versus the rejected subset (or an equivalent calibration diagnostic such as reliability diagrams or ECE) so that readers can verify the assumption that uncertainty is sufficiently well-calibrated.

    Authors: We agree that explicit verification of uncertainty calibration strengthens the central claim. In the revised manuscript we will add reliability diagrams for each soil property and report the observed error (MAE or RMSE) on the accepted versus rejected subsets at the operating points of Figure 4. These diagnostics will be placed in Section 3.3 or a new supplementary section. revision: yes

  2. Referee: [Section 4.1, Table 2] Section 4.1, Table 2: the cost-reduction percentages are computed relative to a baseline of 100 % laboratory measurements. The paper should state the exact per-sample laboratory cost model and the fraction of samples rejected at each operating point; without these numbers the claimed savings cannot be reproduced or compared across laboratories.

    Authors: We accept that the cost model and rejection fractions are required for reproducibility. The revised manuscript will state the per-sample laboratory cost assumption (Québec regional pricing) and add a column or supplementary table listing the exact rejection rate for each property and uncertainty threshold used in Table 2. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper presents an empirical framework for reject-to-remeasure using probabilistic foundation models on a regional soil spectroscopy dataset. The central claim rests on experimental results showing uncertainty-guided rejection meets user accuracy thresholds while reducing costs. No equations, derivations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text. The approach is data-driven and externally benchmarked against conventional lab methods, remaining self-contained without any reduction of outputs to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no free parameters, axioms, or invented entities are identifiable from the provided text.

pith-pipeline@v0.9.1-grok · 5731 in / 997 out tokens · 14041 ms · 2026-06-26T14:29:25.958673+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

59 extracted references · 3 linked inside Pith

  1. [1]

    Science 327(5967), 828–831 (2010)

    Gebbers, R., Adamchuk, V.I.: Precision agriculture and food security. Science 327(5967), 828–831 (2010)

    2010

  2. [2]

    Geoderma292, 59–86 (2017) 19

    Minasny, B., Malone, B.P., McBratney, A.B., Angers, D.A., Arrouays, D., Cham- bers, A., Chaplot, V., Chen, Z.-S., Cheng, K., Das, B.S.,et al.: Soil carbon 4 per mille. Geoderma292, 59–86 (2017) 19

    2017

  3. [3]

    Geoderma 199, 54–63 (2013)

    Brodsk` y, L., Vaˇ s´ at, R., Klement, A., Z´ adorov´ a, T., Jakˇ s´ ık, O.: Uncertainty prop- agation in vnir reflectance spectroscopy soil organic carbon mapping. Geoderma 199, 54–63 (2013)

    2013

  4. [4]

    Geoderma Regional39, 00874 (2024)

    ˇZ´ ıˇ zala, D., Princ, T., Sk´ ala, J., Juˇ ricov´ a, A., Lukas, V., Bohovic, R., Z´ adorov´ a, T., Minaˇ r´ ık, R.: Soil sampling design matters-enhancing the efficiency of digital soil mapping at the field scale. Geoderma Regional39, 00874 (2024)

    2024

  5. [5]

    Agricultural systems148, 71–74 (2016)

    Viscarra Rossel, R.A., Bouma, J.: Soil sensing: A new paradigm for agriculture. Agricultural systems148, 71–74 (2016)

    2016

  6. [6]

    Geoderma137(1-2), 70–82 (2006)

    Viscarra Rossel, R., McGlynn, R., McBratney, A.: Determining the composi- tion of mineral-organic mixes using uv–vis–nir diffuse reflectance spectroscopy. Geoderma137(1-2), 70–82 (2006)

    2006

  7. [7]

    Advances in agronomy132, 139–159 (2015)

    Nocita, M., Stevens, A., Wesemael, B., Aitkenhead, M., Bachmann, M., Barth` es, B., Dor, E.B., Brown, D.J., Clairotte, M., Csorba, A.,et al.: Soil spectroscopy: an alternative to wet chemistry for soil monitoring. Advances in agronomy132, 139–159 (2015)

    2015

  8. [8]

    Journal of Geophysical Research: Solid Earth95(B8), 12653–12680 (1990)

    Clark, R.N., King, T.V., Klejwa, M., Swayze, G.A., Vergo, N.: High spectral resolution reflectance spectroscopy of minerals. Journal of Geophysical Research: Solid Earth95(B8), 12653–12680 (1990)

    1990

  9. [9]

    Soil Science Society of America Journal 59(2), 364–372 (1995)

    Ben-Dor, E., Banin, A.: Near-infrared analysis as a rapid method to simultane- ously evaluate several soil properties. Soil Science Society of America Journal 59(2), 364–372 (1995)

    1995

  10. [10]

    ¨O., Madeno˘ glu, S., Sand´ en, T., Gholizadeh, A., Stenberg, B.,et al.: Accuracies and costs of prediction and mapping soil properties using proximal sensors: A systematic review

    Lozano-Fond´ on, C., Lorenzetti, R., Barbetti, R., Metzger, K., Buttafuoco, G., Pinar, M. ¨O., Madeno˘ glu, S., Sand´ en, T., Gholizadeh, A., Stenberg, B.,et al.: Accuracies and costs of prediction and mapping soil properties using proximal sensors: A systematic review. Computers and Electronics in Agriculture243, 111378 (2026)

    2026

  11. [11]

    Geoderma352, 251–267 (2019)

    Ng, W., Minasny, B., Montazerolghaem, M., Padarian, J., Ferguson, R., Bailey, S., McBratney, A.B.: Convolutional neural network for simultaneous predic- tion of several soil properties using visible/near-infrared, mid-infrared, and their combined spectra. Geoderma352, 251–267 (2019)

    2019

  12. [12]

    Barkov, V., Schmidinger, J., Gebbers, R., Atzmueller, M.: Modern neural net- works for small tabular datasets: The new default for field-scale digital soil mapping? European Journal of Soil Science77(2), 70299 (2026)

    2026

  13. [13]

    Earth-Science Reviews 155, 198–230 (2016) 20

    Viscarra Rossel, R., Behrens, T., Ben-Dor, E., Brown, D., Demattˆ e, J.A.M., Shepherd, K.D., Shi, Z., Stenberg, B., Stevens, A., Adamchuk, V.,et al.: A global spectral library to characterize the world’s soil. Earth-Science Reviews 155, 198–230 (2016) 20

    2016

  14. [14]

    PloS one20(1), 0296545 (2025)

    Safanelli, J.L., Hengl, T., Parente, L.L., Minarik, R., Bloom, D.E., Todd-Brown, K., Gholizadeh, A., Mendes, W.d.S., Sanderman, J.: Open soil spectral library (ossl): Building reproducible soil calibration models through open development and community engagement. PloS one20(1), 0296545 (2025)

    2025

  15. [15]

    Geoderma425, 116029 (2022)

    Poppiel, R.R., Silveira Paiva, A.F., Demattˆ e, J.A.M.: Bridging the gap between soil spectroscopy and traditional laboratory: Insights for routine implementation. Geoderma425, 116029 (2022)

    2022

  16. [16]

    Soil & Environmental Health3(3), 100160 (2025)

    Sharififar, A., Francos, N., Ng, W., Minasny, B., Gholizadeh, A., Karunaratne, S., McBratney, A.: Navigating the challenges of spectral soil sensing: Key solutions for success. Soil & Environmental Health3(3), 100160 (2025)

    2025

  17. [17]

    European Journal of Soil Science73(1), 13192 (2022)

    McBride, M.B.: Estimating soil chemical properties by diffuse reflectance spec- troscopy: Promise versus reality. European Journal of Soil Science73(1), 13192 (2022)

    2022

  18. [18]

    Discussion paper, Saint Loup Research Institute (2022)

    Baveye, P.C.: Limited usefulness of visible-near-infrared spectroscopy in soils: The picture gets much clearer. Discussion paper, Saint Loup Research Institute (2022). Saint Loup Research Institute Discussion Papers, Volume 3, Paper 2. https://www.researchgate.net/publication/366250041

    arXiv 2022

  19. [19]

    Geoderma434, 116491 (2023)

    Santana, F.B., Hall, R.L., Lowe, V., Browne, M.A., Grunsky, E.C., Fitzsimons, M.M., Gallagher, V., Daly, K.: A systematic approach to predicting and map- ping soil particle size distribution from unknown samples using large mid-infrared spectral libraries covering large-scale heterogeneous areas. Geoderma434, 116491 (2023)

    2023

  20. [20]

    Soil science society of America journal66(3), 988–998 (2002)

    Shepherd, K.D., Walsh, M.G.: Development of reflectance spectral libraries for characterization of soil properties. Soil science society of America journal66(3), 988–998 (2002)

    2002

  21. [21]

    Journal of the Royal Statistical Society Series B: Statistical Methodology69(2), 243–268 (2007)

    Gneiting, T., Balabdaoui, F., Raftery, A.E.: Probabilistic forecasts, calibration and sharpness. Journal of the Royal Statistical Society Series B: Statistical Methodology69(2), 243–268 (2007)

    2007

  22. [22]

    Geoderma437, 116585 (2023)

    Schmidinger, J., Heuvelink, G.B.M.: Validation of uncertainty predictions in digital soil mapping. Geoderma437, 116585 (2023)

    2023

  23. [23]

    Geoderma425, 116063 (2022)

    Padarian, J., Minasny, B., McBratney, A.: Assessing the uncertainty of deep learning soil spectral models using monte carlo dropout. Geoderma425, 116063 (2022)

    2022

  24. [24]

    Chemometrics and Intelligent Laboratory Systems265, 105473 (2025) 21

    Wadoux, A.M.-C., Ramirez-Lopez, L.: Uncertainty of predictions in absorp- tion spectroscopy: Modelling with quantile regression forest. Chemometrics and Intelligent Laboratory Systems265, 105473 (2025) 21

    2025

  25. [25]

    Geoderma401, 115155 (2021)

    Wadoux, A.M.J.-C., Heuvelink, G.B.M., Lark, R.M., Lagacherie, P., Bouma, J., Mulder, V.L., Libohova, Z., Yang, L., McBratney, A.B.: Ten challenges for the future of pedometrics. Geoderma401, 115155 (2021)

    2021

  26. [26]

    Spatial Statistics50, 100619 (2022)

    Lark, R., Chagumaira, C., Milne, A.: Decisions, uncertainty and spatial informa- tion. Spatial Statistics50, 100619 (2022)

    2022

  27. [27]

    IEEE Trans

    Chow, C.K.: On optimum recognition error and reject tradeoff. IEEE Trans. Inf. Theory16, 41–46 (1970)

    1970

  28. [28]

    Proceedings of the IEEE111(2), 185–215 (2023)

    Zhang, X.-Y., Xie, G.-S., Li, X., Mei, T., Liu, C.-L.: A survey on learning to reject. Proceedings of the IEEE111(2), 185–215 (2023)

    2023

  29. [29]

    Machine Learning113(5), 3073–3110 (2024)

    Hendrickx, K., Perini, L., Plas, D., Meert, W., Davis, J.: Machine learning with a reject option: A survey. Machine Learning113(5), 3073–3110 (2024)

    2024

  30. [30]

    Journal of biomedical optics16(6), 067009–067009 (2011)

    Rodriguez-Diaz, E., Castanon, D.A., Singh, S.K., Bigio, I.J.: Spectral classi- fier design with ensemble classifiers and misclassification-rejection: application to elastic-scattering spectroscopy for detection of colonic neoplasia. Journal of biomedical optics16(6), 067009–067009 (2011)

    2011

  31. [31]

    NPJ Digital Medicine4(1), 4 (2021)

    Kompa, B., Snoek, J., Beam, A.L.: Second opinion needed: communicating uncertainty in medical machine learning. NPJ Digital Medicine4(1), 4 (2021)

    2021

  32. [32]

    Canadian Journal of soil science89(5), 531–541 (2009)

    Nduwamungu, C., Ziadi, N., Parent, L.- ´E., Tremblay, G.F., Thuri` es, L.: Opportu- nities for, and limitations of, near infrared reflectance spectroscopy applications in soil analysis: A review. Canadian Journal of soil science89(5), 531–541 (2009)

    2009

  33. [33]

    arXiv preprint arXiv:2512.11779 (2025)

    Braun, S., Holzm¨ uller, D., Jordan, M.I., Bach, F.: Conditional coverage diagnos- tics for conformal prediction. arXiv preprint arXiv:2512.11779 (2025)

    Pith/arXiv arXiv 2025

  34. [34]

    Quar- terly Journal of the Royal Meteorological Society: A journal of the atmospheric sciences, applied meteorology and physical oceanography135(643), 1512–1519 (2009)

    Br¨ ocker, J.: Reliability, sufficiency, and the decomposition of proper scores. Quar- terly Journal of the Royal Meteorological Society: A journal of the atmospheric sciences, applied meteorology and physical oceanography135(643), 1512–1519 (2009)

    2009

  35. [35]

    TrAC Trends in Analytical Chemistry135, 116166 (2021)

    Barra, I., Haefele, S.M., Sakrabani, R., Kebede, F.: Soil spectroscopy with the use of chemometrics, machine learning and pre-processing techniques in soil diagnosis: Recent advances–a review. TrAC Trends in Analytical Chemistry135, 116166 (2021)

    2021

  36. [36]

    Soil Science Society of America Journal89(2), 70028 (2025)

    Gozukara, G., Hartemink, A.E., Huang, J., Demattˆ e, J.A.M.: Prediction accuracy of pxrf, mir, and vis-nir spectra for soil properties—a review. Soil Science Society of America Journal89(2), 70028 (2025)

    2025

  37. [37]

    arXiv preprint arXiv:2506.16791 (2025)

    Erickson, N., Purucker, L., Tschalzev, A., Holzm¨ uller, D., Desai, P.M., Salinas, D., Hutter, F.: Tabarena: A living benchmark for machine learning on tabular 22 data. arXiv preprint arXiv:2506.16791 (2025)

    Pith/arXiv arXiv 2025

  38. [38]

    Remote Sensing Applications: Society and Environment, 101815 (2025)

    Bourriz, M., Laamrani, A., Ait Abdelali, H., Bourzeix, F., El Bouanani, N., Chehbouni, A.: Assessing hyperspectral sensor capabilities: Prisma vs enmap for crop type mapping in a semi-arid region. Remote Sensing Applications: Society and Environment, 101815 (2025)

    2025

  39. [39]

    Agronomy11(6), 1028 (2021)

    Tavares, T.R., Molin, J.P., Nunes, L.C., Wei, M.C.F., Krug, F.J., Carvalho, H.W.P., Mouazen, A.M.: Multi-sensor approach for tropical soil fertility analy- sis: comparison of individual and combined performance of vnir, xrf, and libs spectroscopies. Agronomy11(6), 1028 (2021)

    2021

  40. [40]

    Communications in soil science and plant analysis15(12), 1409–1416 (1984)

    Mehlich, A.: Mehlich 3 soil test extractant: A modification of mehlich 2 extractant. Communications in soil science and plant analysis15(12), 1409–1416 (1984)

    1984

  41. [41]

    CRC Press, London (1998)

    Heuvelink, G.B.M.: Error Propagation in Environmental Modelling with GIS. CRC Press, London (1998)

    1998

  42. [42]

    Soil5(2), 275–288 (2019)

    Ellinger, M., Merbach, I., Werban, U., Ließ, M.: Error propagation in spectro- metric functions of soil organic carbon. Soil5(2), 275–288 (2019)

    2019

  43. [43]

    European Journal of Soil Science76(1), 70053 (2025)

    Sanderman, J., Partida, C., Safanelli, J.L., Shepherd, K., Ge, Y., Mitu, S.M., Ferguson, R.: Application of a handheld near infrared spectrophotometer to farm- scale soil carbon monitoring. European Journal of Soil Science76(1), 70053 (2025)

    2025

  44. [44]

    Kock, A.-L., Ramphisa-Nghondzweni, P.D., Van Zijl, G.: Development of soil spectroscopy models for the western highveld region, south africa: Why do we need local data? European Journal of Soil Science75(6), 70014 (2024)

    2024

  45. [45]

    Soil11(2), 553–563 (2025)

    Huang, Y.-C., Padarian, J., Minasny, B., McBratney, A.B.: Using monte carlo conformal prediction to evaluate the uncertainty of deep-learning soil spectral models. Soil11(2), 553–563 (2025)

    2025

  46. [46]

    Qu´ ebec, Canada: Institut de Recherche et de D´ eveloppement en Agroenvironnement (IRDA) (2023)

    Gasser, M.-O., Boss´ e, C., Cl´ ement, C., Bernard, C., Mathieu, J., Tremblay, M.: Rapport 1 de l’´ etude sur l’´ etat de sant´ e des sols agricoles du qu´ ebec:´Etat de sant´ e des principales s´ eries de sols cultiv´ ees. Qu´ ebec, Canada: Institut de Recherche et de D´ eveloppement en Agroenvironnement (IRDA) (2023)

    2023

  47. [47]

    Nature637(8045), 319–326 (2025)

    Hollmann, N., M¨ uller, S., Purucker, L., Krishnakumar, A., K¨ orfer, M., Hoo, S.B., Schirrmeister, R.T., Hutter, F.: Accurate predictions on small data with a tabular foundation model. Nature637(8045), 319–326 (2025)

    2025

  48. [48]

    arXiv preprint arXiv:2511.08667 (2025) 23

    Grinsztajn, L., Fl¨ oge, K., Key, O., Birkel, F., Jund, P., Roof, B., J¨ ager, B., Safaric, D., Alessi, S., Hayler, A., et al.: Tabpfn-2.5: Advancing the state of the art in tabular foundation models. arXiv preprint arXiv:2511.08667 (2025) 23

    Pith/arXiv arXiv 2025

  49. [49]

    arXiv preprint arXiv:2602.11139 (2026)

    Qu, J., Holzm¨ uller, D., Varoquaux, G., Morvan, M.L.: Tabiclv2: A better, faster, scalable, and open tabular foundation model. arXiv preprint arXiv:2602.11139 (2026)

    arXiv 2026

  50. [50]

    Econometrica: journal of the Econometric Society, 33–50 (1978)

    Koenker, R., Bassett Jr, G.: Regression quantiles. Econometrica: journal of the Econometric Society, 33–50 (1978)

    1978

  51. [51]

    III: The collinearity problem in linear regression

    Wold, S., Ruhe, A., Wold, H., Dunn, W.J. III: The collinearity problem in linear regression. the partial least squares (pls) approach to generalized inverses. SIAM Journal on Scientific and Statistical Computing5(3), 735–743 (1984)

    1984

  52. [52]

    Geoderma459, 117337 (2025)

    Schmidinger, J., Vogel, S., Barkov, V., Pham, A.-D., Gebbers, R., Tavakoli, H., Correa, J., Tavares, T.R., Filippi, P., Jones, E.J.,et al.: Limesoda: A dataset collection for benchmarking of machine learning regressors in digital soil mapping. Geoderma459, 117337 (2025)

    2025

  53. [53]

    the Journal of machine Learning research12, 2825–2830 (2011)

    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V.,et al.: Scikit-learn: Machine learning in python. the Journal of machine Learning research12, 2825–2830 (2011)

    2011

  54. [54]

    Journal of machine learning research7(6) (2006)

    Meinshausen, N., Ridgeway, G.: Quantile regression forests. Journal of machine learning research7(6) (2006)

    2006

  55. [55]

    Geoderma291, 55–64 (2017)

    Vaysse, K., Lagacherie, P.: Using quantile regression forest to estimate uncertainty of digital soil mapping products. Geoderma291, 55–64 (2017)

    2017

  56. [56]

    Soil and Tillage Research195, 104410 (2019)

    Szatm´ ari, G., Pirk´ o, B., Ko´ os, S., Laborczi, A., Bakacsi, Z., Szab´ o, J., P´ asztor, L.: Spatio-temporal assessment of topsoil organic carbon stock change in hungary. Soil and Tillage Research195, 104410 (2019)

    2019

  57. [57]

    Geoderma455, 117223 (2025)

    John, K., Saurette, D.D., Heung, B.: The problematic case of data leakage: A case for leave-profile-out cross-validation in 3-dimensional digital soil mapping. Geoderma455, 117223 (2025)

    2025

  58. [58]

    February (2014)

    Stevens, A., Ramirez-Lopez, L.: An introduction to the prospectr package. February (2014)

    2014

  59. [59]

    Group, C.A.S.C.C.S.C.W., Canada, N.R.C., Agriculture, C., Branch, A.-F.C.R.: The Canadian System of Soil Classification vol. 1646. NRC Research Press, Ottawa (1998) Extended Data Figures and Tables 24 Table S1 Predictive performance of the ML models (extended).Probabilistic performance was evaluated using CRPS (continuous ranked probability score; lower m...

    1998