Machine-learning models of &delta;13C and &delta;15N isoscapes in Amazonian wood

Souza-Silva, Isabela M.; Martinelli, Luiz A.; Holmes, Brent; Batista, Ana C. G.; Araújo, Maria G. S.; Garção, Anna L.; Ponton, Stéphane; Groenendijk, Peter; Locosselli, Giuliano M.; Ortega-Rodriguez, Daigard R. O.; Amorim, Deoclecio J.; Costa, Fábio J. V.; Nardoto, Gabriela B.; Brunello, Alexandre T.; Costa, Vladimir Eliodoro; Assis-Pereira, Gabriel; Tomazello-Filho, Mario; Higuchi, Niro; Barbosa, Ana C.; Sena-Souza, João Paulo; Bataille, Clément P.

doi:10.5194/egusphere-2025-5452

Preprints

https://doi.org/10.5194/egusphere-2025-5452

Preprints

13 Nov 2025

| 13 Nov 2025

Machine-learning models of δ¹³C and δ¹⁵N isoscapes in Amazonian wood

Isabela M. Souza-Silva, Luiz A. Martinelli, Brent Holmes, Ana C. G. Batista, Maria G. S. Araújo, Anna L. Garção, Stéphane Ponton, Peter Groenendijk, Giuliano M. Locosselli, Daigard R. O. Ortega-Rodriguez, Deoclecio J. Amorim, Fábio J. V. Costa, Gabriela B. Nardoto, Alexandre T. Brunello, Vladimir Eliodoro Costa, Gabriel Assis-Pereira, Mario Tomazello-Filho, Niro Higuchi, Ana C. Barbosa, João Paulo Sena-Souza, and Clément P. Bataille

Abstract. Illegal logging is one of the most prevalent environmental infractions in the Amazon, led by organized networks that cause substantial ecological and economic impacts. Official control mechanisms, such as Brazil’s Forest Origin Document (DOF), remain vulnerable to the fraudulent manipulation of virtual timber credits and inconsistencies in digital traceability. These deficiencies highlight the need for independent, scientifically based methodologies for timber traceability that can support law enforcement and ensure reliable provenance verification. Here, we tested whether the isotopic composition of carbon (δ¹³C) and nitrogen (δ¹⁵N) in wood can trace Amazonian timber origin. We developed basin-wide δ¹³C and δ¹⁵N isoscapes using machine-learning models to predict spatial variability. A total of 571 trees from 47 sites were analyzed for both isotopes. Tree disks or wedges were sampled from the basal trunk, sectioned transversely, and sub-sampled from heartwood to near the sapwood boundary to obtain whole-tree isotopic composition. The δ¹³C and, more strongly, the δ¹⁵N values exhibited substantial within-site heterogeneity, indicating individual-level physiological controls, interspecific differences, and/or small-scale environmental variation influencing isotope fractionation. Despite these sources of noise, isotopic values showed independent and predictable spatial patterns across the basin (R² = 0.67 for δ¹⁵N and R² = 0.60 for δ¹³C). Nitrogen isotopes were primarily controlled by edaphic factors, while carbon isotopes revealed a broad longitudinal gradient linked to climate. Together, these isotopic markers provide complementary information for basin-scale timber provenancing and form a robust, high-resolution framework for Amazon-wide traceability.

How to cite. Souza-Silva, I. M., Martinelli, L. A., Holmes, B., Batista, A. C. G., Araújo, M. G. S., Garção, A. L., Ponton, S., Groenendijk, P., Locosselli, G. M., Ortega-Rodriguez, D. R. O., Amorim, D. J., Costa, F. J. V., Nardoto, G. B., Brunello, A. T., Costa, V. E., Assis-Pereira, G., Tomazello-Filho, M., Higuchi, N., Barbosa, A. C., Sena-Souza, J. P., and Bataille, C. P.: Machine-learning models of δ¹³C and δ¹⁵N isoscapes in Amazonian wood, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-5452, 2025.

Received: 04 Nov 2025 – Discussion started: 13 Nov 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 5407 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (5407 KB)

Supplement (1596 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

02 Feb 2026

| Highlight paper

Machine-learning models of δ¹³C and δ¹⁵N isoscapes in Amazonian wood

Isabela M. Souza-Silva, Luiz A. Martinelli, Brent Holmes, Ana C. G. Batista, Maria G. S. Araújo, Anna L. Garção, Stéphane Ponton, Peter Groenendijk, Giuliano M. Locosselli, Daigard R. Ortega-Rodriguez, Deoclecio J. Amorim, Fábio J. V. Costa, Gabriela B. Nardoto, Alexandre T. Brunello, Vladimir Eliodoro Costa, Gabriel Assis-Pereira, Mario Tomazello-Filho, Niro Higuchi, Ana C. Barbosa, João Paulo Sena-Souza, and Clément P. Bataille

Biogeosciences, 23, 881–904, https://doi.org/10.5194/bg-23-881-2026,https://doi.org/10.5194/bg-23-881-2026, 2026

Short summary Co-editor-in-chief

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2025-5452', Anonymous Referee #1, 09 Dec 2025
The manuscript describes an extensive sampling of 15N and 13C in wood across the Amazon. A major goal of the project is to investigate whether isotopes can be used to provenance lumber, with relevance to monitoring illegal logging. Towards this goal, random forests models are fitted to the isotope datasets using a large suite of assembled spatial, ecological, pedological, and climatological covariates. RF models are subsequently used to produce isoscapes: spatially resolved isotope predictions across the Amazon. Patterns of isoscape variation are discussed in the context of various ecological factors.

Major comments
Given the collection of two independent isotope signals, I wonder at the missed opportunity here to generate a combined isoscape of 13C and 15N variation. It seems the combined variation in both isotopes would be more spatially variable given distinct controls. Mapping the product of the two, or some other decomposition or principal component, could actually generate a more useful metric.

The study effort, particularly around sample collection, is ambitious and admirable. I agree that the current results are not particularly useful for provenance verification. But perhaps combined with other isotope maps (strontium?) could be in the future.

The machine-learning methods appear robust, but I would be curious to see added some additional discussion around systematic bias of the RF models, which appear to underpredict high values and overpredict low values systematically.

Line 123. Perhaps worth explaining how a sample was collected 215 km from a road, and how frequent such samples are in the dataset. I am imagining this was transported by boat? Then could redefine as distance from road or river?

Fig 1. This figure could be made more useful by somehow depicting sample size at each site, either as point size / color or with a number. This would allow visualization of the distribution of sampling intensity across the region.
L349. You might refer to this as bias. 13C isoscape also exhibits the same bias.
L440. Many sites have 13c variation of around 3-4 per mil. So while I suppose this is true, inter-site variability is not much greater than within site.
Citation: https://doi.org/10.5194/egusphere-2025-5452-RC1
- AC1: 'Reply on RC1', Isabela Maria Souza Silva, 20 Dec 2025
  
  We sincerely thank Referee 1 for the thorough review, thoughtful feedback, and support of our manuscript. Our detailed response letter is attached for your consideration.
  
  Citation: https://doi.org/10.5194/egusphere-2025-5452-AC1
RC2:
'Comment on egusphere-2025-5452', Bin Yang, 16 Dec 2025

General comments
As a stable-isotope researcher with >15 years of experience, I read the work by Souza-Silva et al. on “Machine-learning models of δ¹³C and δ¹⁵N isoscapes in Amazonian wood” with great interest. The authors deliver an exceptionally valuable resource to the community: isotopic measurements for >550 trees distributed across 47 Amazonian sites, coupled with a robust Random-Forest framework that links wood δ¹³C and δ¹⁵N to 74 environmental predictors. The rationale is sound and the methodology state-of-the-art. I therefore strongly recommend its publication. Nevertheless, two issues merit attention before the manuscript reaches its full potential.
Specific comments
1. Basin-scale timber provenancing: although the δ¹³C and δ¹⁵N isoscapes represent a commendable advance, their ability to discriminate Amazonian from non-Amazonian timber (or other regions within the Amazonian) remains equivocal. The observed intra- and inter-site variability is sufficiently large that isotopic signatures from other tropical forests may fall within the Amazonian envelope. I therefore urge the authors to (i) incorporate a comparative dataset of timber collected outside the basin and (ii) provide quantitative classification metrics (sensitivity, specificity) to rigorously demonstrate—and transparently bound—the utility of the isoscapes for large-scale provenance assignment.
2. Conciseness and focus: the manuscript, in its present form, is unduly verbose. Lengthy expositions in the Introduction, Results, and Discussion dilute the key messages and obscure the novel contributions. I strongly recommend a comprehensive tightening of the text—prioritizing only the most essential findings and their implications—to enhance clarity, readability, and impact for the broad readership of this journal.

Citation: https://doi.org/10.5194/egusphere-2025-5452-RC2
- AC2: 'Reply on RC2', Isabela Maria Souza Silva, 20 Dec 2025
  
  We sincerely thank Dr. Bin Yang for his thorough review, insightful comments, and supportive evaluation of our manuscript. Our detailed response letter is attached for your consideration.
  
  Citation: https://doi.org/10.5194/egusphere-2025-5452-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2025-5452', Anonymous Referee #1, 09 Dec 2025
The manuscript describes an extensive sampling of 15N and 13C in wood across the Amazon. A major goal of the project is to investigate whether isotopes can be used to provenance lumber, with relevance to monitoring illegal logging. Towards this goal, random forests models are fitted to the isotope datasets using a large suite of assembled spatial, ecological, pedological, and climatological covariates. RF models are subsequently used to produce isoscapes: spatially resolved isotope predictions across the Amazon. Patterns of isoscape variation are discussed in the context of various ecological factors.

Major comments
Given the collection of two independent isotope signals, I wonder at the missed opportunity here to generate a combined isoscape of 13C and 15N variation. It seems the combined variation in both isotopes would be more spatially variable given distinct controls. Mapping the product of the two, or some other decomposition or principal component, could actually generate a more useful metric.

The study effort, particularly around sample collection, is ambitious and admirable. I agree that the current results are not particularly useful for provenance verification. But perhaps combined with other isotope maps (strontium?) could be in the future.

The machine-learning methods appear robust, but I would be curious to see added some additional discussion around systematic bias of the RF models, which appear to underpredict high values and overpredict low values systematically.

Line 123. Perhaps worth explaining how a sample was collected 215 km from a road, and how frequent such samples are in the dataset. I am imagining this was transported by boat? Then could redefine as distance from road or river?

Fig 1. This figure could be made more useful by somehow depicting sample size at each site, either as point size / color or with a number. This would allow visualization of the distribution of sampling intensity across the region.
L349. You might refer to this as bias. 13C isoscape also exhibits the same bias.
L440. Many sites have 13c variation of around 3-4 per mil. So while I suppose this is true, inter-site variability is not much greater than within site.
Citation: https://doi.org/10.5194/egusphere-2025-5452-RC1
- AC1: 'Reply on RC1', Isabela Maria Souza Silva, 20 Dec 2025
  
  We sincerely thank Referee 1 for the thorough review, thoughtful feedback, and support of our manuscript. Our detailed response letter is attached for your consideration.
  
  Citation: https://doi.org/10.5194/egusphere-2025-5452-AC1
RC2:
'Comment on egusphere-2025-5452', Bin Yang, 16 Dec 2025

General comments
As a stable-isotope researcher with >15 years of experience, I read the work by Souza-Silva et al. on “Machine-learning models of δ¹³C and δ¹⁵N isoscapes in Amazonian wood” with great interest. The authors deliver an exceptionally valuable resource to the community: isotopic measurements for >550 trees distributed across 47 Amazonian sites, coupled with a robust Random-Forest framework that links wood δ¹³C and δ¹⁵N to 74 environmental predictors. The rationale is sound and the methodology state-of-the-art. I therefore strongly recommend its publication. Nevertheless, two issues merit attention before the manuscript reaches its full potential.
Specific comments
1. Basin-scale timber provenancing: although the δ¹³C and δ¹⁵N isoscapes represent a commendable advance, their ability to discriminate Amazonian from non-Amazonian timber (or other regions within the Amazonian) remains equivocal. The observed intra- and inter-site variability is sufficiently large that isotopic signatures from other tropical forests may fall within the Amazonian envelope. I therefore urge the authors to (i) incorporate a comparative dataset of timber collected outside the basin and (ii) provide quantitative classification metrics (sensitivity, specificity) to rigorously demonstrate—and transparently bound—the utility of the isoscapes for large-scale provenance assignment.
2. Conciseness and focus: the manuscript, in its present form, is unduly verbose. Lengthy expositions in the Introduction, Results, and Discussion dilute the key messages and obscure the novel contributions. I strongly recommend a comprehensive tightening of the text—prioritizing only the most essential findings and their implications—to enhance clarity, readability, and impact for the broad readership of this journal.

Citation: https://doi.org/10.5194/egusphere-2025-5452-RC2
- AC2: 'Reply on RC2', Isabela Maria Souza Silva, 20 Dec 2025
  
  We sincerely thank Dr. Bin Yang for his thorough review, insightful comments, and supportive evaluation of our manuscript. Our detailed response letter is attached for your consideration.
  
  Citation: https://doi.org/10.5194/egusphere-2025-5452-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

ED: Publish subject to minor revisions (review by editor) (05 Jan 2026) by David McLagan

AR by Isabela Maria Souza Silva on behalf of the Authors (08 Jan 2026) Author's response Author's tracked changes Manuscript

ED: Publish subject to technical corrections (09 Jan 2026) by David McLagan

Dear authors,
I am happy to proceed with publication after some minor technical changes largely related to the provision of more conservative conclusive statements surrounding the use of these isoscapes in wood traceability. As the authors have acknowledged, the dataset and approach on its own is at best a complementary tool in tracing wood origin, but not as a primary indicator.

"Our results show that these isotopic systems are informative for ecological processes and also hold practical value for timber traceability. Carbon
isotopes primarily reflect broad climatic gradients, whereas nitrogen isotopes capture more localized edaphic processes. Despite substantial intra-site variance, the combination of these isotopes provides a preliminary screening method that can identify potentially suspicious timber batches."
Please adapt this to:
"Our results show that these isotopic systems are informative for ecological processes and potential to play a complementary role in timber traceability. Carbon isotopes primarily reflect broad climatic gradients, whereas nitrogen isotopes capture more localized edaphic processes. Despite substantial intra-site variance, the combination of these isotopes has potential as a preliminary screening method that could identify potentially suspicious timber by batches rather rather than individual trees."

In addition:
"The practical demands of forest governance make the isoscapes presented here a valuable complement to certification systems and sustainable management, directly strengthening efforts to secure a more reliable legal timber trade and to combat illegal logging in the Amazon."
Please adapt to:
"The practical demands of forest governance make the isoscapes presented here a valuable complement to certification systems and sustainable management, and provides a strong foundational approach in efforts to secure a more reliable legal timber trade and to combat illegal logging in the Amazon."

Good luck with finalising the publication.

Best regards,

Hide

AR by Isabela Maria Souza Silva on behalf of the Authors (16 Jan 2026) Author's response Manuscript

Journal article(s) based on this preprint

02 Feb 2026

| Highlight paper

Machine-learning models of δ¹³C and δ¹⁵N isoscapes in Amazonian wood

Isabela M. Souza-Silva, Luiz A. Martinelli, Brent Holmes, Ana C. G. Batista, Maria G. S. Araújo, Anna L. Garção, Stéphane Ponton, Peter Groenendijk, Giuliano M. Locosselli, Daigard R. Ortega-Rodriguez, Deoclecio J. Amorim, Fábio J. V. Costa, Gabriela B. Nardoto, Alexandre T. Brunello, Vladimir Eliodoro Costa, Gabriel Assis-Pereira, Mario Tomazello-Filho, Niro Higuchi, Ana C. Barbosa, João Paulo Sena-Souza, and Clément P. Bataille

Biogeosciences, 23, 881–904, https://doi.org/10.5194/bg-23-881-2026,https://doi.org/10.5194/bg-23-881-2026, 2026

Short summary Co-editor-in-chief

Supplement

https://doi.org/10.5194/egusphere-2025-5452-supplement

Data sets

Research compendium for 'Machine-learning models of δ¹³C and δ¹⁵N isoscapes in Amazonian wood' Isabela M. Souza-Silva and Clément P. Bataille https://osf.io/u5rws/overview

Viewed

Total article views: 486 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
373	79	34	486	50	14	16

HTML: 373
PDF: 79
XML: 34
Total: 486
Supplement: 50
BibTeX: 14
EndNote: 16

Views and downloads (calculated since 13 Nov 2025)

Month	HTML	PDF	XML	Total
Nov 2025	173	22	10	205
Dec 2025	127	32	21	180
Jan 2026	71	24	3	98
Feb 2026	2	1	0	3

Cumulative views and downloads (calculated since 13 Nov 2025)

Month	HTML	PDF	XML	Total
Nov 2025	173	22	10	205
Dec 2025	127	32	21	180
Jan 2026	71	24	3	98
Feb 2026	2	1	0	3

Viewed (geographical distribution)

Total article views: 505 (including HTML, PDF, and XML) Thereof 505 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 08 Feb 2026

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (5407 KB)
Metadata XML

Short summary

Illegal logging is a major environmental concern in the Amazon. We tested whether the isotopic composition of carbon (δ¹³C) and nitrogen (δ¹⁵N) in wood can support timber traceability. Using machine-learning models, we generated basin-wide isoscapes showing that both isotopes capture consistent environmental gradients, providing a scientific basis to improve provenance verification and guide enforcement efforts.


Total:	0
HTML:	0
PDF:	0
XML:	0