Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript addresses an interesting and relevant topic; however, there is a major methodological issue that must be resolved before the manuscript can be properly evaluated.

The authors combined naturalised and invasive species into a single category (“Invasive species”), explaining that these categories are difficult to distinguish. In my opinion, this approach is not methodologically justified. Although the distinction between naturalised and invasive species can sometimes be challenging, these categories represent fundamentally different ecological statuses and should not be merged in the analysis.

This distinction is particularly important in the context of climate change and global warming, which can significantly influence the dynamics of species expansion, establishment, and population decline. Naturalised populations are not necessarily harmful and may remain stable without showing invasive tendencies. In contrast, invasive species are characterised by ecological impact and rapid spread, which makes their separation from merely naturalised species essential for ecological interpretation.

For this reason, I consider this to be a substantial and decisive methodological issue that requires revision. The categories of naturalised and invasive species should be clearly distinguished and analysed separately, and the results and interpretations should be revised accordingly.

Until this methodological issue is addressed, it is difficult to assess the manuscript in detail. Therefore, I would prefer to evaluate the revised version of the manuscript after this key methodological problem has been resolved, and only then provide further comments on other aspects of the study.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript entitled "Determining the invasive status of alien tree species based on their phenological characteristics" presents an analysis of long‑term phenological observations combined with machine‑learning approaches to evaluate whether the invasion status of woody plant species can be predicted from phenological traits. The study relies on a valuable long‑term dataset (1977–2024) collected in the Botanical Garden of Southern Federal University and applies Random Forest models to classify native, introduced, and invasive species using phenological phases and derived temperature metrics.

Overall, the study addresses a relevant and timely topic in invasion ecology and phenology, and the effort required to compile and structure such a long‑term dataset deserves recognition. The integration of phenological monitoring with machine‑learning techniques represents a promising direction for improving the predictive understanding of biological invasions.

Despite the novelty and the potential importance of the work, several methodological, conceptual, and presentation issues require substantial revision before the manuscript can be considered for publication. Concerns arise regarding the ecological interpretation of invasion status, the statistical robustness of the modelling framework, the treatment of temporal autocorrelation, the justification of predictor selection, and the comparison with previous literature on phenology‑based invasion prediction.

Below, detailed comments are provided by manuscript section with reference to the relevant line numbers.

Lines 29–32: The sentence "Alien plant invasions is a global threat to the ecology and economy" contains grammatical issues and should be corrected ("Alien plant invasions are..."). More importantly, the authors should provide a more comprehensive synthesis of the global literature on plant invasions. The introduction relies on only a few references and does not adequately contextualize the study within the broader invasion ecology framework. Key reviews such as Richardson et al. (2000), Pyšek & Richardson (2010), and Blackburn et al. (2011) should be cited when discussing invasion stages and definitions.

Lines 30-31: I think that you should add these two important and recent references as examples to support your sentence: ”Therefore, developing effective measures to monitor and control the spread of alien species is considered a biosecurity priority”. I would like to suggest:

Raffini, F., Bertorelle, G., Biello, R., D’Urso, G., Russo, D., & Bosso, L. (2020). From nucleotides to satellite imagery: Approaches to identify and manage the invasive pathogen Xylella fastidiosa and its insect vectors in Europe. Sustainability, 12(11), 4508.

Rainford, J., Crowe, A., Jones, G., & van den Berg, F. (2020). Early warning systems in biosecurity; translating risk into action in predictive systems for invasive alien species. Emerging Topics in Life Sciences, 4(5), 453-462.

Lines 33–44: The discussion of woody plant encroachment is somewhat confusing and conceptually mixed with biological invasion processes. Woody encroachment is often driven by land‑use change, fire suppression, or grazing pressure rather than by alien species introduction. The authors should clearly distinguish between native woody encroachment; invasion by alien woody species. Without clarifying this distinction, the ecological framing becomes ambiguous.

Lines 45–58: The definitions of introduction, naturalisation, and invasion are presented but lack proper theoretical grounding. The manuscript should explicitly refer to the widely adopted framework proposed by Blackburn et al. (2011) or Richardson et al. (2000), which provide standardized terminology for invasion stages. Currently, the definitions appear simplified and could lead to confusion.

Lines 59–66: The statement that "all naturalised species may become invasive" is overly deterministic and not supported by the majority of invasion ecology literature. Many naturalised species remain benign. The authors should nuance this statement and cite studies discussing invasion debt and transition probabilities.

Lines 59-61: I think that you should add these two important and recent references as examples to support your sentence: ”The position of an alien species along the 'introduction – naturalization – invasion' gradient is subject to change over time, with the potential for all naturalised species to become invasive”. I would like to suggest:

Bertolino, S., Sciandra, C., Bosso, L., Russo, D., Lurz, P. W., & Di Febbraro, M. (2020). Spatially explicit models as tools for implementing effective management strategies for invasive alien mammals. Mammal Review, 50(2), 187-199.

Daly, E. Z., Chabrerie, O., Massol, F., Facon, B., Hess, M. C., Tasiemski, A., ... & Renault, D. (2023). A synthesis of biological invasion hypotheses associated with the introduction–naturalisation–invasion continuum. Oikos, 2023(5), e09645.

Lines 67–75: The link between phenology and invasion success is introduced but remains superficial. A more detailed review of previous work on phenological niche differentiation, extended growing seasons, and phenological plasticity should be included. Important empirical studies demonstrating earlier leaf‑out or longer growing seasons in invasive plants should be discussed.

Lines 76–83: The explanation of proxy metrics (DOY and temperature sums) is too brief. Previous phenological modelling literature has extensively discussed the advantages and limitations of thermal time models. References to Chuine (2000), Schwartz (2013), or related phenology modelling frameworks should be incorporated.

Lines 84–88: The statement about machine learning applications in phenology is vague. The authors should specify which ML methods have previously been used in phenological modelling (e.g., random forests, gradient boosting, neural networks). Without this context, the novelty of the study remains unclear.

Lines 89–95: The research questions are clearly formulated but could be strengthened by explicitly stating testable hypotheses rather than only exploratory questions.

Lines 98–101: The study area description is minimal. The authors should provide additional ecological context such as vegetation composition / soil types / land‑use history / disturbance regime. These factors strongly influence invasion dynamics and should be discussed.

Lines 102–130: The climate description is excessively detailed relative to its analytical relevance. Several numerical climate statistics are presented without explaining how they influence phenological modelling. The section could be substantially shortened while focusing on variables directly relevant to phenology. Additionally, the source and temporal coverage of climate data should be clearly stated.

Lines 131–141: The classification of species raises several concerns. The authors merge naturalised and invasive species into a single "invasive" class due to identification difficulty. This decision significantly weakens the ecological interpretation of the model. Naturalised and invasive species represent distinct ecological stages with different dynamics. Combining them introduces classification noise and may bias model results. The authors should either: maintain three separate classes (introduced / naturalised / invasive) and justify more rigorously why merging is necessary. To move in the introduction section.

Lines 144–163: The species lists are extensive but unnecessary in the main text. These lists would be more appropriate in supplementary material. Their presence interrupts the methodological flow.

Lines 164–171: The reliance on a Soviet‑era phenological protocol is interesting but requires stronger justification. The authors should explain how observer consistency and methodological standardization were maintained over the 47‑year observation period. Long‑term observational datasets often suffer from observer bias and methodological drift.

Lines 172–180: The comparison with the BBCH system is useful but should include explicit mapping of each phenophase to BBCH codes in a clearer format. Currently, the table is difficult to interpret.

Lines 181–187: The conversion of phenological dates into DOY and SAT is standard practice, but the authors do not describe how missing data were handled. Given the long temporal coverage, missing values are expected and should be addressed explicitly. Additionally, the authors should clarify whether the dataset was standardized or normalized before ML modelling.

Lines 188–194: The definition of SAT uses a base temperature of 5°C. This choice should be justified because base temperature strongly influences accumulated heat metrics. Different phenological models often use thresholds between 0°C and 10°C depending on species.

Lines 195–204: The modelling strategy raises several methodological concerns. First, the Random Forest configuration (100 trees and 5 node splits) appears arbitrary. The authors should perform hyperparameter tuning or justify why these parameters were selected. Second, it is unclear whether the dataset was divided into training and testing sets. The use of out‑of‑bag error alone may not be sufficient to evaluate predictive performance. Third, the repeated sampling of 200 growing seasons per class may introduce bias if temporal autocorrelation is present. Phenological observations across years are not independent samples.

Lines 205–215: The description of performance metrics (precision, recall, F1) is standard but unnecessary at this level of detail for a scientific article. Instead, the authors should focus on explaining how these metrics were calculated within the modelling workflow.

Lines 221–234: The threshold used to define collinearity (r > 0.7 or R² > 0.5) is somewhat inconsistent. If R² is used, the corresponding correlation coefficient would be approximately 0.71. The authors should use consistent criteria.

Lines 235–244: The variable reduction from 24 to 6 phenological metrics is not fully justified. PCA loadings or variable importance values should be reported to demonstrate why specific variables were retained.

Lines 246–264: The interpretation of OOB error rates is limited. The authors should compare model performance with baseline classification accuracy (e.g., random assignment) to demonstrate actual predictive value. Additionally, confidence intervals around model metrics are not reported.

Lines 265–271: The claim that reducing predictors did not significantly affect model performance is not statistically tested. Formal comparison between models should be conducted.

Lines 272–285: The comparison between DOY and SAT metrics is interesting but requires stronger statistical validation. The differences in performance appear relatively small.

Lines 286–307: The species‑level classification results are presented in extremely large tables that are difficult to interpret. Aggregated summaries would be more informative. Furthermore, species‑level predictions may be misleading because the model was trained on phenological cycles rather than independent species units.

Lines 308–316: The interpretation of Mean Decrease Accuracy and Gini importance should be approached with caution. Random Forest importance metrics can be biased toward variables with higher variance. Permutation importance or SHAP values would provide more reliable interpretation.

Lines 362–370: The discussion appropriately revisits the research questions but should be reorganized to better connect empirical findings with existing literature.

Lines 371–378: The hypotheses regarding phenological strategies of invasive species are correctly introduced but are not quantitatively tested in the study.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The introduction section is well written, explaining the key facts about invasive plant species and stages of invasion, from introduction till the species become invasive. The comments given regarding the first version of the manuscript are considered, and according to this, the background of the research in the Introduction section is explained well. The research questions and hypothesis are clearly formulated and they are interesting and important to explore.

In Method section, the study area is well described. However, the informations about exclusion of naturalized species from this study should be given. Also, the source for determining the status of alien species (not naturalized, naturalized or invasive) should be given. The list of native, introduced and invasive species is provided as suplementary file, but there is no source (or citation) how the invasiveness status of the mentioned 19 species was assesed. The procedure of the research is well explained.

However, the terminology for the groups of species should be changed, it is confusing. The invasive species are introduced species. When reading the table 2, it is easy to assume that we have total 29 species, 10 non-invasive + 19 invasive. So it is better to use term introduced non-invasive species and invasive species. In this way the both groups will be easier to distinguish, and it will be more clear that there is 29 non-invasive and not naturalized species + 19 invasive species. It is sufficient to mention that in Methods.

Results are clearly presented. However, the more discussion regarding the finding should be done. For example, why Syringa vulgaris had been classified as a native species in  all iterations. What are possible explanations, is there some specific fact regarding its phenology which lead to obtained results, in order to understand better the limitations of model provided. Otherwise, the Discussion is well written, answering the 3 research question of this study. The conclusions are supported by the results. The obtained results showed that the invasive status of some alien tree species can be predicted based on their phenological characteristics using machine learning algorithms, however, the accuracy is moderate and there are the limitations. Hence, this study is valuable contribution and it can help to develop further research in this field. I recommend accepting this paper for publishing after some minor changes (mentioned above) are made.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Well done!

Author Response

Please see the attachment.

Author Response File: Author Response.pdf