Unveiling Key Factors Shaping Forest Interest and Visits: Toward Effective Strategies for Sustainable Forest Use

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Major Comments:

1. Study Context

• Please provide more details about Tokyo's status as having "the world's largest population". What metrics or standards support this claim?

2. Methodological Concerns

• The justification for using LASSO before piecewise SEM is unclear. For example, why not directly apply piecewise SEM, given the established theoretical foundations?
• The theoretical basis for variable relationships in SEM models needs more explicit articulation. For example, what is the theoretical/empirical basis for relationship directions?

3. Documentation Clarity

• Inconsistent terminology between the main text and supplementary materials (e.g., "residence" vs "home") hinders cross-referencing.

 

This case study provides valuable data on urban residents' forest interest and visitation patterns in one of the world's most densely populated areas. However, to maximize the impact of these significant findings, I strongly encourage the authors to strengthen the manuscript by providing more evident theoretical foundations and corresponding methodological design to help readers better understand the relationships between variables and the rationale behind the analytical approach chosen.

Author Response

Comments 1:
1. Study Context
Please provide more details about Tokyo's status as having "the world's largest population". What metrics or standards support this claim?

Response 1:
Thank you for your valuable comment. We have revised the manuscript to clarify the basis for our statement regarding Tokyo's population size. Specifically, we now specify that Tokyo has "the world's largest urban agglomeration by population" and cite the United Nations Population Division's World Urbanization Prospects: The 2018 Revision [18] as the source.

In this report, the United Nations defines urban agglomerations as contiguous urban areas with high population density. According to this definition, Tokyo was the largest urban agglomeration globally in 2018, with approximately 37 million inhabitants. To address potential ambiguities, we have also noted that other definitions of "city size" may result in different rankings.
 
We have incorporated this clarification into the revised manuscript as follows:
(page 3, 1. Introduction, line 106-113)

Tokyo, which has the world's largest urban agglomeration by population (approximately 37 million) according to the United Nations [20]., could provide important insights in an increasingly urbanized world (The term "largest population" refers to the urban agglomeration of Tokyo, which had approximately 37 million inhabitants in 2018, making it the most populous metropolitan area globally according to the United Nations [20]. While Tokyo ranks highest in terms of urban agglomeration, other definitions of city size, such as "city proper" populations, may yield different rankings (e.g., Chongqing in China has the largest city proper population).

We hope this revision sufficiently clarifies our statement and addresses your concern.

Comments 2:
2. Methodological Concerns
The justification for using LASSO before piecewise SEM is unclear. For example, why not directly apply piecewise SEM, given the established theoretical foundations?
The theoretical basis for variable relationships in SEM models needs more explicit articulation. For example, what is the theoretical/empirical basis for relationship directions?

Response 2:
Thank you for your insightful comments. We recognize the importance of providing a clear rationale for using LASSO prior to piecewise SEM, as well as for explicitly articulating the theoretical basis for the relationships specified in our SEM model.

First, we employed LASSO regression as a variable selection method to refine the set of predictors included in the SEM model. Given that SEM allows for a flexible arrangement of explanatory variables, a preliminary selection step was necessary to identify the most influential variables related to Forest Visit and Forest Interest. This step helps to avoid overfitting and improves model interpretability by excluding variables with negligible contributions.

Second, regarding the causal direction of relationships in the SEM model, we structured the model based on a time-sequential approach, wherein past experiences are assumed to influence current behavior. This approach is grounded in behavioral research, where past exposure to natural environments is often linked to current environmental engagement (e.g., outdoor recreation literature, nature-experience studies). Thus, we determined the directionality of paths in accordance with this theoretical perspective.

To address these concerns, we have added the following explanation to the manuscript:
(page 5, 2.3.4. Piecewise SEM Analysis, line 204-212)

Prior to conducting piecewise SEM, we applied LASSO regression to identify the most influential predictors related to Forest Visit and Forest Interest. This selection step was necessary because SEM allows flexible model specification, and an initial variable reduction process helps avoid overfitting while ensuring model interpretability. Furthermore, the directionality of relationships within the SEM model was determined based on a time-sequential framework, wherein past experiences influence present behaviors. This approach is supported by previous studies in outdoor recreation and environmental psychology, which indicate that early-life experiences with nature shape later environmental engagement [29, 30].

Comments 3:
3. Documentation Clarity
Inconsistent terminology between the main text and supplementary materials (e.g., "residence" vs "home") hinders cross-referencing.

Response3:
Thank you for pointing out the inconsistency in terminology between the main text and supplementary materials. To address this issue, we have standardized the terminology throughout the manuscript by unifying "residence" to "home." Additionally, we conducted a thorough review of the entire text to ensure consistency in other terms as well. For example, instances where "housemaker" and "homemaker" were inconsistently used have been corrected to "homemaker" for uniformity.

We appreciate your valuable feedback, which has helped us improve the clarity and coherence of our manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

1. Brief Summary

In the manuscript, the authors focus on factors influencing the interest of urban visitors in forests. The frequency of forest visits is taken as the core metric. They utilize extensive quantitative research (5,000 respondents) using logistic regression (LASSO) and piecewise Structural Equation Modeling  (PSM). The core part of their interest is the sociodemographic, behavioural and environmental variables and their relevance to the interest in forests and forest visitations. The article suggests interesting conclusions related to the significant impact of the „aesthetic influence of walks“ (scenic walks), the role of indirect children in the households (car ownership aspect for transportation) together, and their early exposure to the forest environment. The authors also indicate a negative correlation between age and forest visits, which they compare to similar research in the final suggestion that various local factors might influence the relationship between age and forest visits.  The study offers practical implications for the support and exploitation of urban forests using sound statistical analysis based on a significant number of respondents.

2. General Comments

Relevance and Originality

Urbanisation and urban forestry are appealing, especially regarding the overall worldwide trend of urban concentration and population growth. The manuscript reasonably illustrates how specific variables (transportation capabilities, leisure activity, age) influence the willingness and opportunities to visit forests. Not many scientific articles on the topics pertinent to Japan (Tokio – a large urbanised area) are available. Therefore, the manuscript brings valuable insights and provides a platform for global comparison in like-to-like environments.

Methodological Quality

The authors used LASSO (logistical regression) for initial explanatory variables influencing interest in forests and forest visits as dependent variables for identification. Furthermore, 95% confidence intervals were identified using standard regression analysis (post-LASSO).  For the non-linear models, piecewise structural equation modelling (pSEM) was utilised, where the overall model performance was evaluated using Fisher’s C statistic, and individual models were evaluated using R-square. This combined strategy appears innovative and methodologically sound.

 

The sample size (N=5000) is relatively high, but the respondent panel is CAWI (web questionnaire) with inherent limitations. Authors, however, correctly mention limitations (internet penetration and age category)

 

The two intrinsically differentiated categories, Forest Interest and Forest Visit, are clearly described. For both, a separate analysis is made with consequent pSEM synthesis. This approach provides reasonable fundaments for conclusions.

Statistical Analysis and Reproducibility

 

The presented results (ROC-AUC, PR-AUC) suggest solid model performance.

 

Authors correctly distinguish between LASSO variable selections and consequent standard regression analysis which increases the transparency of the results

 

The replicability could be improved by including a supplementary file with a more detailed cross-validation lambda value assessment.

 

Interpretation of Results

 

Authors consider the frequency of "Scenic Walks " during holidays as the most influential factor determining the relationship to the forests. This idea is well supported by the data. This is further developed with the implication that increasing opportunities for contact with natural environments through childhood can enhance both interest in and visits to forests.

 

The role of parenting appeared indirectly linked to interest in forests and forest visits via vehicle ownership, which was associated with transportation accessibility. This further transitively leads to the conclusion that public transportation access availability to the urban forest areas might increase interest in forests and forest visits. Also, the synergy between early-age access to forest and transportation availability is noted.

 

The age demographics variable presents controversy in some other studies, including studies from Japan and Korea. Authors assume different urban areas, local factors, natural space availability, and societal norms regarding outdoor activities could be the explanation. For a better understanding of the context and clarity of controversy in Japan, the methodology and context comparison of Morita et al. [27] ([26) in Manuscript factual list of references) The research highlights specific differences that could be described in more detail.

 

3. Structure and Clarity

The manuscript is correctly separated into the parts Introduction, Materials and Methods, Results, Discussion and Conclusion – overall, the text appears clear with a logical flow.

 

 

4. Specific Comments

 

Cited reference numbers seem shifted for example Morita et al. [27] ([26) in Manuscript factual list of references) 

 

Line 32-47  multifunctional context could be extended for a broader perspective for example

 

• Cesaro, L., & Gatto, P. (2008).The multifunctional role of forests: Policies, methods and case studies. D. Pettenella (Ed.). European Forest Institute.

 

• Winkel, G., Lovric, M., Muys, B., Katila, P., Lundhede, T., Pecurul, M., ... & Wunder, S. (2022). Governing Europe's forests for multiple ecosystem services: Opportunities, challenges, and policy options.Forest Policy and Economics,145, 102849.

 

 

Discussion, line. 295–356: Well done on the synthesis in the Discussion part. The questionnaire was administered in January – it is worth mentioning potential limitations that the survey took part in winter, so the seasonal „mood“ bias could have had an influence.

 

5. Ethical and Formal Aspects

Ethical aspects appear to be correctly covered.

Citations and list of references appear to be balanced and correctly presented (with minor

Plagiarism: Not detected

 

6. Overall Recommendation (Visible to Editors Only)

The article has sound data fundament and uses methods consistent with the research objective.

 „Accept after Minor Revisions“ – No significant methodological objections found, the article brings new and valuable information about factors influencing forest interest and forest visit interest in urbanized society

 

7. Recommendations for Improvement (If Revisions Are Needed)

 

• For broader context, please assume an extension of the multifunctional forest function  perspective for example
• Cesaro, L., & Gatto, P. (2008).The multifunctional role of forests: Policies, methods and case studies. D. Pettenella (Ed.). European Forest Institute.
• Winkel, G., Lovric, M., Muys, B., Katila, P., Lundhede, T., Pecurul, M., ... & Wunder, S. (2022). Governing Europe's forests for multiple ecosystem services: Opportunities, challenges, and policy options.Forest Policy and Economics,145, 102849.
• Potential limitations for elevated methodological consistency could be mentioned, such as the fact that the survey took place in winter, so the seasonal „mood“bias could have had an influence.
• For better clarity, please assume unification of the terminology „visit/visitation“
• For assured replicability, please assess by including a supplementary file with more details pertinent to cross-validation lambda value assessment.
• For consistency please check the numbering of references/citations in the text versus list

 

 

8. Overall assessment

Overall, the article is highly valuable, combining the extensive data set, up-to-date statistical means and practical implications for the sustainable usage of urban forests.

Minor improvements relate mainly to methodological limits and their detailed disclosure, bringing a broader context of the topic framing and supplementary file replicability improvement.

Author Response

Comments 1:
Specific Comments
Cited reference numbers seem shifted for example Morita et al. [27] ([26) in Manuscript factual list of references)

Response 1:
Thank you for bringing this issue to our attention. We have corrected the misaligned reference numbers, including the discrepancy in Morita et al. [27] ([26] in the manuscript's factual list of references). Additionally, we have carefully reviewed and updated all reference numbers to ensure accuracy, particularly considering the references added during the revision process.
We appreciate your thorough review and constructive feedback.

Comments 2:
Line 32-47  multifunctional context could be extended for a broader perspective for example
Cesaro, L., & Gatto, P. (2008).The multifunctional role of forests: Policies, methods and case studies. D. Pettenella (Ed.). European Forest Institute.
Winkel, G., Lovric, M., Muys, B., Katila, P., Lundhede, T., Pecurul, M., ... & Wunder, S. (2022). Governing Europe's forests for multiple ecosystem services: Opportunities, challenges, and policy options.Forest Policy and Economics,145, 102849.

For broader context, please assume an extension of the multifunctional forest function perspective for example

Response 2:
Thank you for your valuable suggestion regarding the expansion of the discussion on the multifunctional role of forests. In response, we have revised the manuscript to provide a broader perspective on forest multifunctionality by incorporating additional discussions on its various dimensions, including biological, ecological, functional, and managerial aspects.

Specifically, we have added the following paragraph (Lines 35-44) to acknowledge the historical development and political significance of the concept, concerning the supply and demand of forest ecosystem services:

" The definition of forest multifunctionality, which is deeply embedded in the nature of forests, has been proposed over the years from various perspectives, including biological, ecological, functional, and managerial aspects. In 1992, the United Nations Conference on Environment and Development positioned the multiple ecological, economic, social, and cultural roles of forests at the core of the definition of the Principles of Sustainable Forest Management, leading to the idea of multifunctionality gaining new political momentum [3]. Nevertheless, there remains a mismatch between the increasing demand for diverse forest ecosystem services and the ability of forest owners and managers to profit from their supply [4]. To address this gap, numerous studies have been conducted to explore potential solutions."

Comments 3:
The questionnaire was administered in January – it is worth mentioning potential limitations that the survey took part in winter, so the seasonal „mood“ bias could have had an influence.

Potential limitations for elevated methodological consistency could be mentioned, such as the fact that the survey took place in winter, so the seasonal „mood“bias could have had an influence.

Response 3:
Thank you for your valuable suggestion. We agree that seasonal factors, particularly the fact that the survey was conducted in winter, could have influenced respondents' subjective well-being and mood, potentially affecting their responses. To address this concern, we have added a discussion of this limitation in the manuscript. Specifically, we have now acknowledged the potential impact of seasonal variation and referenced relevant prior research (Connolly, 2013; Feddersen et al., 2016; Keller et al., 2005; Denissen et al., 2008) that highlights the influence of weather and seasonal factors on subjective well-being and mood.

We have incorporated this discussion in the Limitations section of the revised manuscript (page 14, lines 429-439), as follows:

“This study has some limitations that should be acknowledged. One potential limitation is the influence of seasonal factors on survey responses. The survey was con-ducted in winter, and previous studies have shown that weather conditions can significantly impact subjective well-being [31, 32] and mood [33, 34]. Colder temperatures and shorter daylight hours may affect respondents' psychological states and their willingness to engage in outdoor activities. Therefore, the observed relationships be-tween forest interest, forest visits, and other variables might partially reflect seasonal variations rather than stable behavioral patterns. Future research should consider conducting surveys across different seasons to assess the robustness of the findings and to account for potential seasonal biases."

We hope this revision sufficiently addresses your concern.

 

Comments 4:
For better clarity, please assume unification of the terminology „visit/visitation“

Response 4:
Thank you for your helpful suggestion. To ensure consistency and clarity throughout the manuscript, we have unified the terminology by replacing "visitation" with "visit" in all instances.

We appreciate your careful review and believe this modification improves the readability of our paper.

 

Comments 5
For assured replicability, please assess by including a supplementary file with more details pertinent to cross-validation lambda value assessment.

The replicability could be improved by including a supplementary file with a more detailed cross-validation lambda value assessment.

Response 5:
Thank you for your valuable feedback. To enhance replicability, we have included additional details on the cross-validation lambda value assessment in the supplementary material. Specifically, we now provide both the mean binomial deviance and its standard deviation across folds for each lambda value. This additional information will allow readers to better assess the variability and stability of the model selection process.
Furthermore, we have added the following statement to the main text (page 4, line 180-182):

To ensure reproducibility, we recorded both the mean binomial deviance and its standard deviation across folds for each lambda value in the supplementary material (table S2, S3).

 

Comments 6:
For consistency please check the numbering of references/citations in the text versus list

Response 6:
Thank you for pointing out the inconsistency in the numbering of references. Upon reviewing the reference list, we found that one citation was missing, which caused a misalignment in the numbering. We have now added the missing reference at position 11 in the list and have adjusted all reference numbers accordingly throughout the manuscript. This ensures consistency between in-text citations and the reference list. Additionally, we have carefully reviewed the reference list again to ensure accuracy, as new references were added during the revision process.

 

Comments 7:
The age demographics variable presents controversy in some other studies, including studies from Japan and Korea. Authors assume different urban areas, local factors, natural space availability, and societal norms regarding outdoor activities could be the explanation. For a better understanding of the context and clarity of controversy in Japan, the methodology and context comparison of Morita et al. [27]. The research highlights specific differences that could be described in more detail.

Response 7:
Thank you for your valuable comment. We recognize that the relationship between age demographics and forest visitation varies across different studies, including Morita et al. [27]. In response to your suggestion, we have added a discussion comparing our findings with those of Morita et al. and explaining the potential reasons for the observed differences.

The following revisions have been made in the Discussion section (page 13, lines 377-384):

For example, Nagoya City, which was the study site in Morita et al.'s research, is a major city in Japan, just like the Tokyo Special Wards that we surveyed. However, there is a significant difference in the rate of car ownership per household between the two cities. As of 2024, the average number of cars per household was 0.324 in Tokyo’s 23 Wards, whereas it was 0.901 in Nagoya City [34]. This factor could influence accessibility to forested areas, particularly for elderly individuals who may rely more on private vehicles. These findings suggest that existing theories on elderly forest visits may need to be refined by incorporating the role of transportation accessibility in urban contexts.

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

The research presented in the manuscript is interesting. However, there are a few recommendations for a better understanding.

The abstract is expected to find key research results rather than general recommendations. 

Table 4 presents the accuracy of the models, but the authors lack to present which are the models. The abbreviations of the models is not clear.

Table 7 – degrees of freedom I assume

The results of pSEM models are expected to be presented in detail.

Discussions is expected to respond to the 2 objectives presented in lines 99-101. How are the results responding to the objectives? Which strategies were explored? (this was objective no 2)

Section 4.2 – lines 358-359 and the following need clarification – if not included then why are these discussed?

What are the limitations of the research? What about the theoretical and practical implications?

Author Response

Comments 1:
The abstract is expected to find key research results rather than general recommendations.

Response 1:
Thank you for your valuable comments. We acknowledge your concern that the abstract should focus more on key research results rather than general recommendations. In response, we have revised the abstract to highlight the main findings of our analysis, including specific statistical results, while reducing broad policy recommendations.

Specifically, we have incorporated key quantitative results from our piecewise SEM analysis, such as:

Nature experiences in current travel destinations (scenic walks) significantly influenced both forest interest (standardized path coefficient = 0.19) and forest visits (0.30).
Childhood nature experiences and local walking habits were found to have a significant impact on current forest interest and visitation.
The lack of private vehicle ownership (-0.13) and increasing age (-0.21) were identified as negative factors for forest visits, contrasting with prior studies suggesting that older individuals tend to visit natural areas more frequently. We discuss how Tokyo’s low car ownership rate may contribute to this difference.
These revisions ensure that the abstract presents concrete research findings while still maintaining a connection to broader implications. We appreciate your suggestion, which has helped improve the clarity and focus of the abstract.

Comments 2:
Table 4 presents the accuracy of the models, but the authors lack to present which are the models. The abbreviations of the models is not clear.

Response 2:
Thank you for your valuable feedback. To clarify the models used in Table 4, we have revised the text to avoid ambiguity by explicitly describing the process of variable selection through LASSO regression without referring to "models."

Additionally, we have added a supplementary table (Table S6 LASSO-Selected Variables and Coefficients for Forest Interest and Forest Visit) to provide a detailed list of the variables selected by LASSO and their coefficients. We believe this addition enhances the transparency of our methodology.

We appreciate your insightful comments, which have helped improve the clarity of our manuscript (page 6, line 249-252).

before
Through LASSO regression, 43 variables were preliminarily selected for the model with Forest Interest as the dependent variable, and 48 variables were selected for the Forest Visit model. Among these, 28 variables were selected in both models.

after
Through LASSO regression, 43 variables were preliminarily selected when using Forest Interest as the dependent variable, and 48 variables were selected when using Forest Visit as the dependent variable. Among these, 28 variables were common to both.

 

Comments 3
Table 7 – degrees of freedom I assume

Response 3:
Thank you for your careful review. We have revised the label in Table 7 from "freedom" to "degrees of freedom" to ensure clarity and accuracy. We appreciate your feedback.

Response 4:

Thank you for your insightful comment. To provide a more detailed presentation of the pSEM results, we have revised the text to explicitly include the standardized path coefficients (β values) and their significance levels for each model. Specifically, we now state (page 10, line 303-306):

 

"Among all models, 'Scenic Walk in current travel destinations' showed the strongest standardized path coefficient (β) for the dependent variable (β = 0.19, p < 0.01 in Model M1; β = 0.30, p < 0.01 in Model M2; β = 0.30, p < 0.01 in Model M3a; β = 0.18, p < 0.001 in Model M3b), indicating its strong association with both Forest Interest and Forest Visit."

This revision enhances clarity and ensures that key results are explicitly stated within the main text. We appreciate your feedback, which helped us improve the presentation of our findings.

 

Comments 5:

Discussions is expected to respond to the 2 objectives presented in lines 99-101. How are the results responding to the objectives? Which strategies were explored? (this was objective no 2)

 

Response 5:

Thank you for your valuable comment. In response, we have revised the introduction of the Discussion section to explicitly state how our findings address the two research objectives presented in lines 99-101. Further policy and practical implications are provided in the Conclusion section. These changes ensure a clearer alignment between our discussion and the research objectives.

 

Specifically, we have added the following statement on Page 12, Lines 327-332:
"The first objective, factors influencing the interest in and visit to forests, is discussed, focusing on variables that worked well in each analysis. The second objective, strategies to encourage urban residents to visit forests and increase their interest in forests, is discussed based on these findings, particularly in relation to accessibility, early-life nature contact, and cultural infrastructure. Further implications for policy and practice are summarized in the conclusion."

Comments 6:

Section 4.2 – lines 358-359 and the following need clarification – if not included then why are these discussed?

Response 6:

Thank you for your valuable comment. We understand that the rationale for discussing variables not included in the pSEM needed further clarification. To address this, we have revised the introduction to Section 4.2 to explicitly state that while these variables were not part of the pSEM models, they were identified as significant in the post-LASSO analysis. Given their relevance to understanding behavioral patterns and policy implications, we have provided a complementary discussion in this section. This revision clarifies the purpose of including this discussion and aligns it with the overall research framework.

We appreciate your feedback, which has helped improve the clarity of our manuscript (page 13, line 398-401).

before

Although not included in the pSEM, we now discuss variables identified in the post-LASSO analysis as having significant effects on forest interest and visitation.

after

While the following variables were not directly included in the pSEM models, the post-LASSO analysis identified them as having significant effects on forest interest and visitation. Given their potential relevance to understanding behavioral patterns and policy implications, we provide a complementary discussion of these factors in this section.

Comments 7:

What are the limitations of the research? What about the theoretical and practical implications?

Response 7:

Thank you for your insightful comment.

In response, we have added Section 4.3 to explicitly discuss the limitations of our study (page 14, line 429-439).

 

To address theoretical implications, we have added a discussion in Section 4.1.3, focusing on the most significant difference between our findings and previous research.

Specifically, previous studies (e.g., Morita et al.) have reported that older adults are more likely to visit forests. However, our study found the opposite trend, where elderly individuals in Tokyo’s 23 Wards were less likely to visit forests. To explain this discrepancy, we considered differences in transportation accessibility. For instance, the average number of cars per household is significantly lower in Tokyo (0.324) compared to Nagoya (0.901), where Morita et al.’s study was conducted. Given that elderly individuals may rely more on private vehicles, urban environments with lower car ownership rates may pose accessibility barriers to forest visits.

Specifically, we have added the following statement on Page 13, Lines 377-384:

"For example, Nagoya City, which was the study site in Morita et al.'s research, is a major city in Japan, just like the Tokyo Special Wards that we surveyed. However, there is a significant difference in the rate of car ownership per household between the two cities. As of 2024, the average number of cars per household was 0.324 in Tokyo’s 23 Wards, whereas it was 0.901 in Nagoya City [34]. This factor could influence accessibility to forested areas, particularly for elderly individuals who may rely more on private vehicles. These findings suggest that existing theories on elderly forest visits may need to be refined by incorporating the role of transportation accessibility in urban contexts."

By incorporating this discussion, we highlight the theoretical implication that existing theories on elderly forest visitation may need to be refined by considering the role of transportation accessibility in urban settings. We believe this addition strengthens our discussion and provides a clearer theoretical contribution.

 

For practical implications, we have revised the Conclusion section to further highlight the practical implications of our findings. Specifically, we have strengthened the discussion on concrete policy recommendations, emphasizing the importance of accessibility, early-life nature contact, and cultural infrastructure in promoting forest interest and visits. Additionally, we have underscored the broader implications for urban planning, environmental education, and social policies, while also suggesting future research directions to explore the implementation of these strategies in diverse socio-cultural contexts. We believe these revisions enhance the clarity and applicability of our findings.

Specifically, we have revised the text in two sections: Page 14, Lines 441-454 and Lines 455-458. The updated text is as follows:

"The findings from both the pSEM and post-LASSO analyses provide concrete policy implications for promoting forest interest and visits. Given the significant influence of accessibility, early-life nature contact, and cultural infrastructure, targeted interventions should prioritize the following strategies:

"These insights underscore the need for a multifaceted approach that integrates urban planning, environmental education, and social policies to foster stronger connections between people and forests. Future research should explore how these strategies can be effectively implemented in different socio-cultural contexts."

These revisions enhance the discussion of policy implications and strategic recommendations, aligning them more clearly with the study’s findings and research objectives.

We appreciate your valuable feedback, which has helped us refine our argument.

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I appreciate your response regarding the use of LASSO prior to SEM. While your explanation provides some rationale, two important concerns remain:

1. Trade-off in Variable Selection
   • While LASSO can indeed reduce overfitting and improve statistical precision by eliminating variables with negligible statistical contributions, this approach may simultaneously reduce the comprehensive consideration of theoretically important variables.
   • Variables that are conceptually important from a theoretical perspective but have weaker statistical relationships might be excluded, potentially leading to an incomplete understanding of the phenomenon under study.
2. Methodological Precedent
   • Could you please provide specific references to existing studies that have employed this sequential LASSO-then-SEM approach in similar research contexts?
   • Such precedents would help validate your methodological choice and demonstrate its acceptance in the field.

Your time-sequential approach to determining causal directionality is reasonable, but a more explicit discussion of how the LASSO results informed your SEM specification would strengthen the manuscript. This would include explaining which theoretically important variables might have been excluded by LASSO and how you reconciled statistical selection with theoretical considerations.

Author Response

Comment 1:

1. Trade-off in Variable Selection

While LASSO can indeed reduce overfitting and improve statistical precision by eliminating variables with negligible statistical contributions, this approach may simultaneously reduce the comprehensive consideration of theoretically important variables.

Variables that are conceptually important from a theoretical perspective but have weaker statistical relationships might be excluded, potentially leading to an incomplete understanding of the phenomenon under study.

Your time-sequential approach to determining causal directionality is reasonable, but a more explicit discussion of how the LASSO results informed your SEM specification would strengthen the manuscript. This would include explaining which theoretically important variables might have been excluded by LASSO and how you reconciled statistical selection with theoretical considerations.

Response 2:

Thank you for your insightful comment. We agree that LASSO-based selection may overlook theoretically important variables, and we have taken steps to address this issue in our revised manuscript.

In response, we have provided specific examples of variables that were excluded by LASSO but were manually reintegrated into the SEM model based on theoretical considerations. One such variable was Nature experiences around home when under 18 years old: observation, which improved model fit upon reintegration, confirming its theoretical importance. On the other hand, gender, a variable known to affect forest visit (with women visiting less frequently), was excluded by LASSO and did not improve model fit when added back into the SEM model. This suggests that gender may have a less significant effect in our study context than has been reported in prior research. These details have been incorporated into the Method (Page 5, Lines 207-210), Results (Page 11, Lines 322-333) and Limitations (Page 15, Lines 465-479) sections of the manuscript. We hope this provides greater clarity on how variable reintegration informed our final model.

method: 2.3.4 Piecewise SEM Analysis (Page 5, Lines 207-210)

To address the concern regarding potentially omitted variables, we manually reintegrated certain theoretically important variables that were excluded by LASSO. These variables were systematically added to the SEM model, and we monitored changes in model fit indices to assess their impact.

 

Results: 3.4. Results of Piecewise Structural Equation Modeling (pSEM)

(Page 11, Lines 322-333)

Several variables that were excluded by LASSO due to weaker statistical relationships, but were considered theoretically significant, were manually added to the SEM model. Nature experiences around home when under 18 years old: observation, for instance, was not selected by LASSO but was manually added based on its theoretical relevance in understanding forest visits. After adding this variable, the model's fit indices showed improvement. This improvement supports the importance of early nature experiences in influencing later forest visit.

Conversely, gender, which prior studies have identified as a significant factor influencing forest visits (with women visit less frequently [33, 34]), was excluded by LASSO and did not improve the model fit when added to the SEM. This suggests that in our study, the effect of gender on forest visit may be less pronounced, or other variables may better capture this relationship.

 

4.3. Limitations (Page 15, Lines 465-479)

This study employs a sequential LASSO-then-SEM approach, which offers a structured method for variable selection and causal analysis. However, this approach remains relatively novel in applied research, and its implementation in related fields is still emerging. While LASSO effectively reduces dimensionality and prevents overfitting, its tendency to exclude variables with weaker statistical signals—even when they are theoretically important—raises concerns about potential information loss. In this study, we addressed this issue by manually reintegrating certain variables and assessing their im-pact on model fit.

One potential extension of this work is the incorporation of more systematic sensitivity analyses (e.g., stepwise reintegration of excluded variables, Bayesian model averaging) to further assess the robustness of variable selection. However, given the complexity of our SEM framework and the need to balance model parsimony with interpretability, implementing exhaustive sensitivity analyses for all excluded variables was beyond the scope of this study. Future research could explore these methods to refine the LASSO-SEM approach and establish best practices for its application in various domains.

Comments 2:

Methodological Precedent

Could you please provide specific references to existing studies that have employed this sequential LASSO-then-SEM approach in similar research contexts?

Such precedents would help validate your methodological choice and demonstrate its acceptance in the field.

 

Response 2:

Thank you for your valuable feedback. To address your comment, we have now included references to prior studies that have employed the sequential LASSO-then-SEM approach to analyze complex causal mechanisms in different research domains. Specifically:

Domínguez-Miranda, S. A., Rodriguez-Aguilar, R., & Velazquez Salazar, M. (2025). Modeling the Relation Between Non-Communicable Diseases and the Health Habits of the Mexican Working Population: A Hybrid Modeling Approach. Mathematics, 13(6), 959.

Kumagai et al. (2024) applied LASSO to identify key environmental factors influencing black soybean yield and subsequently used SEM to investigate their interdependencies.

These studies provide methodological precedent for our approach and demonstrate its applicability in uncovering complex relationships in diverse research contexts. We have incorporated this discussion in Lines 210–218 on Page 5 of the revised manuscript.

We appreciate your insightful suggestion and believe that this addition strengthens the justification for our methodological framework. Please let us know if further clarification is needed.

method: 2.3.4 Piecewise SEM Analysis (Page 5, Lines 210-218)

The sequential LASSO-then-SEM approach has been employed in prior research to analyze complex causal mechanisms across different domains. For instance, Domínguez-Miranda et al. (2025) integrated LASSO-based variable selection with SEM to examine the relationships between health habits and non-communicable diseases. Similarly, Kumagai et al. (2024) applied LASSO to identify key environmental factors influencing black soybean yield and subsequently used SEM to investigate their interdependencies. These studies provide methodological precedent for our approach, demonstrating its applicability in uncovering complex relationships in diverse research contexts [29, 30].

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript was improved based on recommendations. For an easier way to follow all the changes I recommend in future to mark in text with red or use track changes for all changes! 

Author Response

Comments 1:
The manuscript was improved based on recommendations. For an easier way to follow all the changes I recommend in future to mark in text with red or use track changes for all changes! 

Response 1:
Thank you for your valuable feedback and for recognizing the improvements in the manuscript. In future submissions, we will ensure that changes are clearly marked using track changes or a different color for better readability.