PCLM-D-25-00419

A Hybrid Machine Learning Framework for Land Use Carbon Accounting Identifies Cropland to Forest Conversion as the Dominant Mitigation Lever in Tanzania

PLOS Climate

Dear Dr. Johansen,

Thank you for submitting your manuscript to PLOS Climate. After careful consideration, we feel that it has merit but does not fully meet PLOS Climate’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Apr 09 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at climate@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pclm/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

We look forward to receiving your revised manuscript.

Kind regards,

Jason Morgan

Staff Editor

PLOS Climate

Journal Requirements:

1. Please note that PLOS Climate has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/climate/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

2. We ask that a manuscript source file is provided at Revision. Please upload your manuscript file as a .doc, .docx, .rtf or .tex.

3. Please provide separate figure files in .tif or .eps format.

For more information about figure files please see our guidelines:  LINK

https://journals.plos.org/globalpublichealth/s/figures

https://journals.plos.org/globalpublichealth/s/figures#loc-file-requirements

If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

Additional Editor Comments (if provided):

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

-->Comments to the Author

1. Does this manuscript meet PLOS Climate’s publication criteria? Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe methodologically and ethically rigorous research with conclusions that are appropriately drawn based on the data presented.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->2. Has the statistical analysis been performed appropriately and rigorously?-->

Reviewer #1: Yes

Reviewer #2: No

**********

-->3. Have the authors made all data underlying the findings in their manuscript fully available (please refer to the Data Availability Statement at the start of the manuscript PDF file)?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception. The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: Yes

Reviewer #2: No

**********

-->4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS Climate does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)-->

Reviewer #1: The manuscript addresses a timely topic related to the quantitative assessment of CO2 emissions driven by land-use change, using a hybrid approach that combines econometric models, time-series analysis, and machine learning methods. The study is based on an extensive temporal dataset (1996–2021), employs official national inventory data (IPCC Tier 2), and demonstrates technically sound implementations of MLR, ARIMA/ARIMAX, Random Forest, and XGBoost. However, despite its methodological richness and emphasized policy relevance, the manuscript suffers from methodological redundancy, potential overfitting, limited conceptual novelty, and overly strong interpretative claims that are not fully supported by the data.

The claimed scientific novelty is largely declarative. The use of hybrid models (MLR + ARIMA + RF + XGBoost) is not new in itself, and the conclusion that cropland-to-forest conversion represents the dominant mitigation lever is well established in the literature and reiterates widely accepted views on the role of reforestation in the carbon balance.

The exceptionally high R2 values (≈0.97) obtained for the regression model with a large number of categorical predictors raise serious concerns regarding overfitting and pseudo-determinism. The manuscript lacks an analysis of residual autocorrelation, coefficient stability, and model sensitivity to specification choices.

The MLR framework includes dozens of dummy variables representing land-use transition types, which renders coefficient interpretation highly problematic and calls into question the generalizability of the results beyond the analyzed period.

The ARIMAX model is acknowledged by the authors to perform poorly; however, this result is not examined from a methodological perspective. The manuscript does not discuss why the inclusion of exogenous socio-economic variables degrades model performance or what implications this has for the structural interpretation of emissions.

The superior performance of XGBoost in terms of RMSE and MAE is expected, yet it is not accompanied by an analysis of model robustness, temporal feature importance, or the risks associated with black-box interpretation. In its current form, the machine-learning component appears more as a demonstration of computational capability than as a source of new process-level understanding.

The scenario analysis is based on simple linear modifications of input variables and does not account for institutional, social, or spatial constraints on the implementation of such scenarios, rendering the results illustrative rather than robustly predictive.

The spatial dimension is entirely absent. National-level aggregation smooths out regional contrasts, yet the conclusions are presented as if they were directly applicable to practical land-use management, which is methodologically problematic.

The Discussion section largely reiterates the results and contains broad, generalized statements on the roles of urbanization, GDP, and forests that are already well documented in the literature and do not constitute an independent scientific contribution.

The authors are encouraged to place greater emphasis on more recent studies related to greenhouse gas emissions, including:

https://doi.org/10.1016/j.eiar.2025.107956

https://doi.org/10.1016/j.scitotenv.2024.173895

https://doi.org/10.1016/j.apenergy.2024.124946

https://doi.org/10.1016/j.jclepro.2022.132312

Reviewer #2: Manuscript: “A Hybrid Machine Learning Framework for Land Use Carbon Accounting Identifies Cropland to Forest Conversion as the Dominant Mitigation Lever in Tanzania” (PCLM-D-25-00419)

This manuscript addresses a high priority problem for land sector climate policy and land use governance, namely how to quantify and forecast land use carbon emissions in a context where inventories and monitoring systems are often incomplete, and how to compare mitigation levers such as reforestation with broader socio economic drivers. The attempt to integrate accounting style estimates with regression, time series forecasting, machine learning prediction, and scenario based sensitivity metrics is potentially publishable and could be useful to both scientific and practitioner audiences. However, the current version contains methodological ambiguities and internal inconsistencies that significantly weaken interpretability, credibility, and reproducibility. The most important weaknesses relate to unclear definition of the unit of analysis across the different modelling components, a lack of alignment between claimed preprocessing steps and how coefficients are interpreted, and a validation design for machine learning that appears vulnerable to time leakage and therefore likely inflates predictive performance. As a result, the strength of the headline conclusion about a dominant mitigation lever and the interpretation of socio economic effects are not yet sufficiently supported.

Contribution framing and novelty

The manuscript would benefit from a more precise articulation of its scientific contribution. The hybrid framework and the LUCSI concept are presented as novel, yet the paper does not clearly demonstrate how the approach differs from established land use accounting workflows supplemented by standard predictive models and a subsequent scenario analysis. A stronger framing would explicitly separate methodological innovation from a Tanzania application case study, and from benchmarking of common algorithms. The introduction should also set clear boundaries around what the framework can and cannot infer, especially where the manuscript makes statements that imply decision readiness. In its current form, the novelty claim reads broader than what the methods and validation can justify.

Data structure and carbon accounting workflow

A major issue requiring immediate clarification is the apparent mismatch between the dataset description and the implied modelling sample size. The manuscript describes annual data spanning 1996 to 2021, which would ordinarily imply a short national time series. Yet the reported regression degrees of freedom indicate a much larger number of observations, which strongly suggests that the authors constructed a stacked or panel style dataset such as transition by year or land use category by year. That transformation may be appropriate, but it is not explained, and the reader is left unable to determine what one row represents in the regression and machine learning models. Because land use carbon accounting depends on definitions of land categories, transitions, activity data, and emission factors, the paper must describe a transparent workflow. It should explain how emissions are computed from activity data and Tier 2 factors, how land use types and transitions are defined and encoded, whether transitions are mutually exclusive and exhaustive, how land balance constraints are handled, and what the modelling unit is for each component including regression, time series modelling, and machine learning. Without these details, neither inference nor forecasts can be trusted.

Preprocessing transformations and internal consistency

The manuscript states that missing values were imputed using MICE, continuous variables were log transformed, and categorical land use variables were dummy encoded. Yet the interpretation of coefficients in the results and discussion does not align with those statements. Interpreting a coefficient as a direct additive change in emissions per percentage point change in a predictor is not straightforward when log transformations are used, unless the authors explicitly show the functional form and provide correct back transformations. More critically, there is an internal contradiction between the sign and magnitude of the urbanization effect reported in the narrative and the coefficient shown in the results table. This is not a minor wording issue because it reverses the direction of a key socio economic conclusion. The manuscript must reconcile the transformation pipeline and confirm whether coefficients are presented on the original scale or the transformed scale. It must also rewrite interpretations using appropriate elasticity language where relevant and report uncertainty in a way consistent with the model specification.

Regression specification and inference validity

The regression includes a large number of land use category or transition predictors alongside socio economic covariates. If land use predictors represent proportions or compositional shares, then dependence among predictors is inevitable because shares sum to one, and this can cause multicollinearity and unstable coefficient estimates unless handled carefully with a reference category and appropriate compositional treatment. If the authors constructed a stacked dataset with multiple transition observations per year, then socio economic covariates may repeat across rows for the same year, violating independence assumptions and leading to overly optimistic standard errors unless clustering or other robust approaches are used. The very high R squared reported could reflect explanatory power, but it could also arise because land use indicators encode structural elements already embedded in the accounting computation, which would make the regression partially tautological rather than independent. The manuscript should explicitly identify the reference category, justify the inclusion of specific land use terms, report multicollinearity checks, provide residual diagnostics, and explain how variance estimation accounts for heteroscedasticity and potential within year dependence.

Time series modelling alignment with the data generating process

The ARIMA and ARIMAX components are described as if the response variable is a single annual national emissions series. That is reasonable if the objective is to forecast national totals. However, if the regression and machine learning models operate on a stacked dataset with many more observations, then the paper must explain what series is being modelled in the time series section and how that target is made comparable to the machine learning prediction task. The manuscript mentions rolling origin evaluation but does not provide enough detail for reproducibility. It should specify the forecast horizon, the training window, the origin points, the number of folds, and the evaluation metrics. If model comparisons are retained, the comparison must be between models trained and evaluated on equivalent targets and validation protocols.

Machine learning design validation and leakage control

The machine learning section reports an eighty twenty split and also refers to cross validation. If the underlying data are time indexed, random splitting and standard k fold validation can leak information from later years into training, inflating performance and overstating generalization. For forecasting and policy planning, time aware validation is essential. The manuscript should implement a strict chronological split and preferably rolling origin evaluation for machine learning models as well. Preprocessing steps including imputation and transformations must also be performed within each training fold to prevent leakage through the preprocessing pipeline. The large performance differences reported across algorithms warrant stronger reporting on reproducibility. The paper should confirm identical partitions and feature sets, report tuning procedures clearly, and provide uncertainty around performance metrics through repeated runs or resampling.

Scenario analysis and LUCSI construction

The scenario analysis and LUCSI concept could provide decision relevant marginal insights, but the method requires tighter specification. If LUCSI is a derivative based sensitivity, the manuscript must explain precisely how derivatives are obtained from models that encode land use transitions using dummy variables rather than continuous shares, and how units are preserved across any transformations. Scenario construction also needs explicit feasibility and accounting constraints. A scenario that reallocates a large fraction of cropland to forest must respect land balance and plausible transition rates, and the paper should demonstrate how those constraints are implemented. Scenario results should include uncertainty bounds. Point estimates alone are insufficient for policy claims because they hide the sensitivity of conclusions to model error and data uncertainty.

Interpretation and policy claims

The discussion sometimes uses causal language when the study design as presented supports association rather than causal identification. Even if accounting identities imply mechanistic differences among land transitions, the socio-economic modelling does not by itself establish causality without stronger identification strategies. The manuscript should therefore tighten its language and separate accounting-based statements from statistical associations. It should also moderate claims about dominance or decision finality until validation and uncertainty analysis support such strong conclusions. Once internal inconsistencies are resolved and validation is strengthened, the paper can still offer meaningful prioritization insights, but those insights should be framed as evidence informed and conditional rather than definitive.

Reproducibility and presentation quality

Given the complexity of imputation, transformations, stacked data construction, model tuning, and scenario generation, the data and code availability approach of providing materials upon request is not sufficient for reproducibility. The manuscript should provide a public repository containing the analysis pipeline and, where possible, a processed dataset or a transparent procedure for rebuilding it from raw sources. A complete variable dictionary with units and transformation notes is also necessary. Presentation issues must be corrected, particularly contradictions between narrative statements and tables, inconsistent terminology for land use types versus transitions, and inconsistent reporting of evaluation metrics across modelling approaches.

**********

-->6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

Do you want your identity to be public for this peer review? If you choose “no”, your identity will remain anonymous but your review may still be made public.

For information about this choice, including consent withdrawal, please see our Privacy Policy.-->

Reviewer #1: No

Reviewer #2: Yes: Isaac Tchuwa

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

--> -->-->Figure Resubmissions:

-->-->While revising your submission, we strongly recommend that you use PLOS’s NAAS tool (https://ngplosjournals.pagemajik.ai/artanalysis) to test your figure files. NAAS can convert your figure files to the TIFF file type and meet basic requirements (such as print size, resolution), or provide you with a report on issues that do not meet our requirements and that NAAS cannot fix.-->-->

After uploading your figures to PLOS’s NAAS tool - https://ngplosjournals.pagemajik.ai/artanalysis, NAAS will process the files provided and display the results in the "Uploaded Files" section of the page as the processing is complete. If the uploaded figures meet our requirements (or NAAS is able to fix the files to meet our requirements), the figure will be marked as "fixed" above. If NAAS is unable to fix the files, a red "failed" label will appear above. When NAAS has confirmed that the figure files meet our requirements, please download the file via the download option, and include these NAAS processed figure files when submitting your revised manuscript.-->

Attachments

Attachment

Submitted filename: Peer Review Report.pdf

PCLM-D-25-00419R1

A Hybrid Machine Learning Framework for Land Use Carbon Accounting Identifies Cropland to Forest Conversion as the Dominant Mitigation Lever in Tanzania

PLOS Climate

Dear Dr. Johansen,

Thank you for submitting your manuscript to PLOS Climate. After careful consideration, we feel that it has merit but does not fully meet PLOS Climate’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The reviewers have provided their comments below. Please pay particular attention to their feedback about the discussion of the machine learning experiments.

Please submit your revised manuscript by May 22 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at climate@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pclm/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

As the corresponding author, your ORCID iD is verified in the submission system and will appear in the published article. PLOS supports the use of ORCID, and we encourage all coauthors to register for an ORCID iD and use it as well. Please encourage your coauthors to verify their ORCID iD within the submission system before final acceptance, as unverified ORCID iDs will not appear in the published article. Only the individual author can complete the verification step; PLOS staff cannot verify ORCID iDs on behalf of authors.

We look forward to receiving your revised manuscript.

Kind regards,

Joanna Tindall, PhD

Staff Editor

PLOS Climate

Journal Requirements:

If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

Additional Editor Comments (if provided):

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

-->Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.-->

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

**********

-->2. Does this manuscript meet PLOS Climate’s publication criteria? Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe methodologically and ethically rigorous research with conclusions that are appropriately drawn based on the data presented.-->

Reviewer #1: Partly

Reviewer #2: Yes

**********

-->3. Has the statistical analysis been performed appropriately and rigorously?-->

Reviewer #1: N/A

Reviewer #2: Yes

**********

-->4. Have the authors made all data underlying the findings in their manuscript fully available (please refer to the Data Availability Statement at the start of the manuscript PDF file)?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception. The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS Climate does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)-->

Reviewer #1: While the revised manuscript shows clear improvement, the major concerns have been acknowledged but not fully resolved, with several issues addressed primarily at the level of interpretation rather than methodology.

The regression model still relies on land-use transition variables that are structurally embedded in the emissions accounting framework. Although the authors now recognize this, the model specification remains unchanged, and the high R2 should be interpreted with greater caution as it partly reflects accounting consistency rather than independent explanatory power.

The clarification of the stacked dataset is helpful; however, the repetition of socio-economic variables within each year may still affect statistical inference. While clustered standard errors are applied, the implications of this data structure (e.g., effective sample size and dependence) are not fully discussed.

The machine learning component has been more cautiously framed, but the validation design remains relatively weak. The use of a single temporal split does not fully eliminate concerns about temporal leakage, and therefore the reported predictive performance should be interpreted carefully.

Finally, the scenario analysis has been improved, but it remains based on stylized input perturbations and should be clearly presented as illustrative rather than policy-predictive.

Reviewer #2: The revised manuscript has improved substantially and addresses most of the major methodological and conceptual concerns raised in the previous review. The authors have made commendable efforts to strengthen transparency, clarify the analytical framework, and moderate earlier overstatements. As a result, the manuscript is now more coherent, reproducible, and methodologically defensible. However, a few points could still be strengthened before publication.

First, the manuscript should be more cautious in how it presents the machine learning results. Although the use of a chronological train–test split is an improvement, the text should acknowledge more explicitly that the validation remains relatively limited and that predictive performance should therefore be interpreted with some caution.

Second, the manuscript would benefit from clearer wording on the comparability of the modelling approaches. Since the ARIMA models are fitted to aggregated national time series while the machine learning models use the stacked dataset, the paper should avoid making overly direct comparisons across model classes and should clarify this limitation in the discussion.

Third, the scenario analysis could be strengthened by explaining the assumptions behind the land-use transition scenarios more clearly. In particular, the manuscript should briefly justify why the selected scenarios are reasonable and relevant, even if they are intended primarily as illustrative policy scenarios.

Overall, this is a much improved revision. The remaining issues are relatively minor and can be addressed through clearer framing, more cautious interpretation, and slightly more explicit explanation of assumptions.

**********

-->7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

Do you want your identity to be public for this peer review? If you choose “no”, your identity will remain anonymous but your review may still be made public.

For information about this choice, including consent withdrawal, please see our Privacy Policy.-->

Reviewer #1: No

Reviewer #2: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

--> -->-->Figure Resubmissions:

-->-->While revising your submission, we strongly recommend that you use PLOS’s NAAS tool (https://ngplosjournals.pagemajik.ai/artanalysis) to test your figure files. NAAS can convert your figure files to the TIFF file type and meet basic requirements (such as print size, resolution), or provide you with a report on issues that do not meet our requirements and that NAAS cannot fix.-->-->

After uploading your figures to PLOS’s NAAS tool - https://ngplosjournals.pagemajik.ai/artanalysis, NAAS will process the files provided and display the results in the "Uploaded Files" section of the page as the processing is complete. If the uploaded figures meet our requirements (or NAAS is able to fix the files to meet our requirements), the figure will be marked as "fixed" above. If NAAS is unable to fix the files, a red "failed" label will appear above. When NAAS has confirmed that the figure files meet our requirements, please download the file via the download option, and include these NAAS processed figure files when submitting your revised manuscript.-->

Attachments

Attachment

Submitted filename: Reviewer Report-PLCM-25-00419_R1.pdf

PCLM-D-25-00419R2

A Hybrid Machine Learning Framework for Land Use Carbon Accounting Identifies Cropland to Forest Conversion as the Dominant Mitigation Lever in Tanzania

PLOS Climate

Dear Dr. Johansen,

Thank you for submitting your manuscript to PLOS Climate. After careful consideration, we feel that it has merit but does not fully meet PLOS Climate’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The manuscript has been evaluated by two reviewers, and their comments are available below. The reviewers have raised some additional concerns with your manuscript that need attention. Could you please revise the manuscript to carefully address the concerns raised?

Please submit your revised manuscript by Jun 28 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at climate@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pclm/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

As the corresponding author, your ORCID iD is verified in the submission system and will appear in the published article. PLOS supports the use of ORCID, and we encourage all coauthors to register for an ORCID iD and use it as well. Please encourage your coauthors to verify their ORCID iD within the submission system before final acceptance, as unverified ORCID iDs will not appear in the published article. Only the individual author can complete the verification step; PLOS staff cannot verify ORCID iDs on behalf of authors.

We look forward to receiving your revised manuscript.

Kind regards,

Brian Patrick Weaver, Ph.D.

Staff Editor

PLOS Climate

Journal Requirements:

If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

Additional Editor Comments (if provided):

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

-->Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.-->

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

-->2. Does this manuscript meet PLOS Climate’s publication criteria? Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe methodologically and ethically rigorous research with conclusions that are appropriately drawn based on the data presented.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->3. Has the statistical analysis been performed appropriately and rigorously?-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->4. Have the authors made all data underlying the findings in their manuscript fully available (please refer to the Data Availability Statement at the start of the manuscript PDF file)?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception. The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS Climate does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)-->

Reviewer #1: Thank you for the detailed revision of your manuscript PCLM-D-25-00419_R2 and for the thorough point-by-point response. I appreciate the substantial effort you have made to clarify methodological limitations, moderate your claims, and improve overall transparency. The revised version is certainly stronger and more honest about the conditional nature of your findings.

However, before final acceptance, I would like to suggest three minor but essential revisions that would further align the manuscript's presentation with its actual analytical rigor.

1. Revise the title for accuracy.

The current title states that the framework “Identifies Cropland to Forest Conversion as the Dominant Mitigation Lever.” Given your repeated and appropriate caveats throughout the text (e.g., “model based associations rather than causal dominance,” “conditional on model specification,” “not causal inference”), the word “Identifies” is too strong. I recommend a more cautious formulation, for example:

“A Hybrid Machine Learning Framework Suggests Cropland to Forest Conversion as a Potential Mitigation Lever in Tanzania”

or simply “A Hybrid Machine Learning Framework for Land Use Carbon Accounting: A Case Study of Tanzania.” This change would eliminate the tension between your title and your internal disclaimers.

2. Substantially revise the LUCSI presentation or remove it.

The Land Use Carbon Sensitivity Index (LUCSI) remains poorly defined for categorical land use transitions. The current definition LUCSIk=ΔCO2/ΔXk does not specify how ΔXk is measured for a transition such as “Cropland → Forest.” Reporting LUCSI values in tCO2/USDt implies a level of precision and economic interpretation that is not justified by the data or model structure. If you wish to keep LUCSI as an exploratory heuristic, please provide a clear, step by step numerical example of its calculation for at least one transition. Otherwise, I suggest removing LUCSI entirely from the main text and placing it in an appendix as an experimental indicator.

3. Tone down the policy recommendations in Section 4.2.

Your recommendations (e.g., “targeted land use policies should be implemented”) read as strong normative prescriptions, while your own analysis is explicitly conditional and scenario based. Please rephrase these recommendations to reflect their exploratory and conditional nature. For instance: “If the observed associations are confirmed in future research, then promoting reforestation and cropland to forest conversion could be a promising mitigation strategy” rather than direct policy mandates.

These revisions are modest in scope but important for scientific consistency. I look forward to seeing the final corrected version.

Reviewer #2: The authors have largely addressed the issues raised in the previous review. The revised manuscript is clearer, more balanced, and more cautious in its interpretation of the findings. The concerns regarding machine-learning validation, comparability of modelling approaches, and scenario assumptions have been substantially addressed.

Only minor issues remain. These relate mainly to final wording and presentation. The authors may still wish to slightly moderate any remaining strong claims about XGBoost performance, ensure that model comparisons remain properly qualified, and undertake final language editing to correct grammar, spacing, and formatting issues.

**********

-->7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

Do you want your identity to be public for this peer review? If you choose “no”, your identity will remain anonymous but your review may still be made public.

For information about this choice, including consent withdrawal, please see our Privacy Policy.-->

Reviewer #1: No

Reviewer #2: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

--> -->-->Figure Resubmissions:

-->-->While revising your submission, we strongly recommend that you use PLOS’s NAAS tool (https://ngplosjournals.pagemajik.ai/artanalysis) to test your figure files. NAAS can convert your figure files to the TIFF file type and meet basic requirements (such as print size, resolution), or provide you with a report on issues that do not meet our requirements and that NAAS cannot fix.-->-->

After uploading your figures to PLOS’s NAAS tool - https://ngplosjournals.pagemajik.ai/artanalysis, NAAS will process the files provided and display the results in the "Uploaded Files" section of the page as the processing is complete. If the uploaded figures meet our requirements (or NAAS is able to fix the files to meet our requirements), the figure will be marked as "fixed" above. If NAAS is unable to fix the files, a red "failed" label will appear above. When NAAS has confirmed that the figure files meet our requirements, please download the file via the download option, and include these NAAS processed figure files when submitting your revised manuscript.-->

Attachments

Attachment

Submitted filename: Review Report-PCLM-D-25-00419_R2.pdf

A Hybrid Machine Learning Framework for Land Use Carbon Accounting: A Case Study of Tanzania

PCLM-D-25-00419R3

Dear Mr Johansen,

We are pleased to inform you that your manuscript 'A Hybrid Machine Learning Framework for Land Use Carbon Accounting: A Case Study of Tanzania' has been provisionally accepted for publication in PLOS Climate.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow-up email from a member of our team.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they'll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact climate@plos.org.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Climate.

Best regards,

Jamie Males

Staff Editor

PLOS Climate

***********************************************************

Additional Editor Comments (if provided):

Reviewer Comments (if any, and for reference):

Reviewer's Responses to Questions

-->Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.-->

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

-->2. Does this manuscript meet PLOS Climate’s publication criteria? Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe methodologically and ethically rigorous research with conclusions that are appropriately drawn based on the data presented.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->3. Has the statistical analysis been performed appropriately and rigorously?-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->4. Have the authors made all data underlying the findings in their manuscript fully available (please refer to the Data Availability Statement at the start of the manuscript PDF file)?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception. The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS Climate does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)-->

Reviewer #1: Thanks to the authors for the work done, I recommend this article

Reviewer #2: The revised manuscript has adequately addressed the comments. The authors have moderated the interpretation of the machine-learning results, clarified the limited comparability between ARIMA and machine-learning models, and improved the explanation of the scenario assumptions.

The manuscript is now clearer, more cautious, and methodologically better framed. Any remaining issues are minor/editorial and do not affect the overall contribution of the work.

**********

-->7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

Do you want your identity to be public for this peer review? If you choose “no”, your identity will remain anonymous but your review may still be made public.

For information about this choice, including consent withdrawal, please see our Privacy Policy.-->

Reviewer #1: No

Reviewer #2: No

**********

Attachments

Attachment

Submitted filename: Review_Report-PLCM-25-00419_R3.pdf