Review Reports - Towards Sustainable Waste Management: Predictive Modelling of Illegal Dumping Risk Zones Using Circular Data Loops and Remote Sensing

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript introduces a novel and comprehensive framework for identifying and managing illegal waste dumping sites, centered on an innovative "circular data loop" concept. The topic is forward-looking, applying a sophisticated integration of GIS analysis, remote sensing, and machine learning to a persistent environmental and public health problem. The iterative nature of the model is a significant methodological strength and highly relevant to the journal's scope of promoting sustainable, data-driven management practices.

While the study presents a compelling approach with valuable contributions, several areas require further refinement and clarification before it can be considered for publication. My comments for improvement are detailed below:

Major Concerns:

Lack of Quantitative Model Validation: The study's central claim is that each iteration of the data loop improves the prediction of risk areas. This is compellingly illustrated in Figure 16, which shows a reduction in the total surface area classified as high-risk. However, this is an indirect and qualitative measure of performance. The manuscript lacks standard quantitative validation metrics for the machine learning models. To robustly prove that the model is improving, the authors should present metrics such as Area Under the Curve (AUC) for a Receiver Operating Characteristic (ROC) curve, or Precision-Recall curves. For instance, the newly identified LNOP sites in Loop 2 and Loop 3 could be used as a test set to evaluate the predictive power of the model from the previous loop. Without this quantitative assessment, the claim of "improved prediction" remains suggestive rather than rigorously demonstrated.
Discussion on Generalizability and Transferability:The study is an excellent case study focused on the Municipality of Maribor, Slovenia. However, the manuscript would be significantly strengthened by a more thorough discussion of its generalizability. The authors should dedicate a subsection in the "Discussion" to the challenges and requirements for transferring this model to other study areas. Key questions to address include: What are the minimum required datasets for this framework to be effective? How might the influential factors (and their weights in the model) differ in regions with different characteristics? How scalable is the EkoVaruh application and the overall workflow? A discussion on these points would elevate the paper from a local case study to a more broadly applicable methodological contribution.

Minor Concerns:

Role of UAV Technology:The use of UAVs for identifying new LNOPs is a modern and effective technique, as demonstrated. However, the role of the UAV should be clarified: is it primarily a validation tool or a discovery tool? A more nuanced discussion on the strategic deployment of UAVs would add significant practical value.
Clarity of Figures and Captions: While the figures are generally informative, some could be improved: (1)The caption for Figure 13 contains a typo: "Figure 13Figure 8 displays...". This should be corrected. (2)The caption for Figure 16, the most critical result, could be more descriptive. It should explicitly state the key takeaway. (3)The maps (e.g., Figures 7, 9, 11) could benefit from a clearer legend or context, especially for readers unfamiliar with the geography of Maribor.

Author Response

We would like to humbly and gratefully thank you for your professional, constructive review of the article. We have considered all your comments, suggestions and guidance. The changes in the article are shown below as before/after. Once again, many thanks.

Changes and comments of the re-submitted paper "Modelling of risk areas with potential for illegal dumping: integrating databases, GIS analysis and remote sensing into a circular data loop"

MAJOR CONCERNS:

Lack of Quantitative Model Validation: The study's central claim is that each iteration of the data loop improves the prediction of risk areas. This is compellingly illustrated in Figure 16, which shows a reduction in the total surface area classified as high-risk. However, this is an indirect and qualitative measure of performance. The manuscript lacks standard quantitative validation metrics for the machine learning models. To robustly prove that the model is improving, the authors should present metrics such as Area Under the Curve (AUC) for a Receiver Operating Characteristic (ROC) curve, or Precision-Recall curves. For instance, the newly identified LNOP sites in Loop 2 and Loop 3 could be used as a test set to evaluate the predictive power of the model from the previous loop. Without this quantitative assessment, the claim of "improved prediction" remains suggestive rather than rigorously demonstrated.

Author comment: More quantitative results explanations were added. Moreover, the reviewer’s request for some additional metrics (e.g., AUC, ROC) was addressed by explaining that the PU dataset and anomaly detection methods prevent the use of traditional metrics. A new Figure was added and a Table 4 with model accuracy. Figure 16 is now Figure 17.

Before	Re-submitted paper
Figure 17: illustrates the surface area of risk zones (in km²), as defined by the risk assessment model, in relation to the number of known LNOP sites recorded in each iteration. The analysis of the graph indicates an inverse relationship between the number of registered LNOP locations and the total area of modelled high-risk zones. When interpreting these results, it is important to consider that the first and second loops are spatially comparable, as the second field campaign included a verification of previously known locations from the first loop, along with the identification of additional sites detected during field control. This means that the second survey was geographically focused and targeted. In contrast, the third loop shows the most significant improvement in the model. Despite including the highest number of recorded locations, the surface area of risk-classified areas is considerably reduced (red bar). This is a result of the location-unbiased data collection approach, which did not rely on prior knowledge of existing sites. The third iteration included random field inspections conducted as part of the "Cleanup Campaign Maribor 2025 – My Waste, My Responsibility” campaign and a systematically distributed quadrant-based survey executed during student field exercises. These findings confirm that a greater volume of high-quality data enables more stringent and effective modelling, leading to more precise targeting of monitoring and remediation activities.	Table 3 and Figure 16 present the recorded illegal waste dumping sites (LNOP) from the perspective of their individual volume and distribution across volume classes. A comparison between loops reveals a very distinct trend of increasing numbers of smaller-volume LNOPs, while the number of larger-volume LNOPs has grown only marginally. This indicates that the most high-risk and environmentally burdensome locations were already identified in the initial phase of assessment, whereas the subsequently identified smaller-volume LNOPs are mostly the result of later illegal activities and, naturally, more detailed field inspections. Figure 16: Number of LNOP by volume class (m³) in each individual loop Figure 17: illustrates the surface area of risk zones (in km²), as defined by the risk assessment model, in relation to the number of known LNOP sites recorded in each iteration. The analysis of the graph shows a direct relationship between the number of identified LNOPs and the area assessed as having no risk of LNOPs occurrence, as well as with the area classified as having the highest probability of occurrence. In contrast, an inverse relationship is observed for areas with a lower probability of occurrence. Such distribution indicates an improvement in results with each successive iteration, as the exclusion of non-risk areas and the clear delineation of high-risk zones are crucial for the usability of data in subsequent field identification of new LNOPs. The observed trend reflects precisely this refinement — a clearer definition of the extreme classification levels (complete absence of LNOP and highest likelihood of occurrence). The results also confirm that a greater volume of high-quality data enables more stringent and effective modelling, which in turn allows for more precise targeting of spatial monitoring activities and subsequent remediation.
	… After Figure 17: The iterative application of the circular data loop through three phases (Loop 1–3) demonstrated clear improvements in both data quality and spatial precision: LNOP locations were increased from 150 to 463 locations, estimated waste volume has increased from 1,310 m³ to 2,827 m³ and the area with no risk and the area with the highest level of risk were very clearly defined. As additional confirmation of the approach, the prediction model’s performance was verified for each iteration using the available LNOP data per loop, as shown in Table 4. The table illustrates that each subsequent model yields improved prediction accuracy — most notably, the third loop model achieves the highest accuracy when tested on the latest dataset (84.5%) as well as consistently high values across earlier iterations. Table 4: Model accuracy based on training and test data across individual loops
	Added to chapter “4. Discussion”: … Validation of the prediction model also presents a significant challenge due to the nature of the available data and the algorithms used. Since the dataset follows a positive-unlabeled (PU) structure—where only known illegal dumping sites are labeled and all other locations remain unlabeled—the absence of confirmed negative examples makes it difficult to objectively evaluate model accuracy using traditional validation methods (e.g., precision, recall, ROC curves). The algorithms employed (One-Class SVM, Isolation Forest, and the Elkan-Noto method) are specifically designed for scenarios where true negatives are unknown, but this also limits the possibility of robust cross-validation or ground truth comparison. Consequently, any assessment of model performance must rely on indirect evidence, expert interpretation, or future detection of new LNOP cases in predicted high-risk areas, which can then serve as post hoc validation. This methodological constraint is inherent to many real-world anomaly detection problems and remains a subject of ongoing research. …

Table 3: Number of LNOP by volume class (m³) in each individual loop, including the increase in number between loops within each class

Volume class (m3)	Loop 1	Difference between Loop 1 and Loop 2	Loop 2	Difference between Loop 2 and Loop 3	Loop 3
0 - 1	47	35	82	87	169
1 - 3	16	19	35	20	55
3 - 9	51	60	111	50	161
9 - 25	24	24	48	4	52
25 - 50	8	9	17	2	19
50 - 75	4	2	6	1	7

Figure 16: Number of LNOP by volume class (m³) in each individual loop

New Figure 17

Figure 17: Comparison of risk area model results by data loop. The bar plot shows overlap of high-risk areas across three models: 0% (green) are areas flagged by none, while 33% (blue), 66% (orange), and 100% (red) correspond to areas predicted risky by one, two, and all three models, respectively.

Table 4: Model accuracy based on training and test data across individual loops

	Validation data
	Loop 1	Loop 2	Loop 3
Training data - Loop 1	64.7	59.9	63.9
Training data - Loop 2	74.0	80.9	82.7
Training data - Loop 3	74.0	79.9	84.5

Discussion on Generalizability and Transferability: The study is an excellent case study focused on the Municipality of Maribor, Slovenia. However, the manuscript would be significantly strengthened by a more thorough discussion of its generalizability. The authors should dedicate a subsection in the "Discussion" to the challenges and requirements for transferring this model to other study areas. Key questions to address include: What are the minimum required datasets for this framework to be effective? How might the influential factors (and their weights in the model) differ in regions with different characteristics? How scalable is the EkoVaruh application and the overall workflow? A discussion on these points would elevate the paper from a local case study to a more broadly applicable methodological contribution.

Author comment: Workflow and EkoVaruh application were described more in a scalable sense. Discussion section has been extended and updated with more related work comparison, some explanation about model improvement, potential about public sector implementation and regional engagement, challenges and requirements for transferring the model to other study areas. Defining a minimum required dataset is challenging, as every selected feature contributed meaningful information, and we utilized all available datasets to maximize model performance. Several features were chosen based on their successful use in related studies, while others were selected through expert judgment, considering their relevance to illegal dumping behavior.

More changes are described in further comments. More references were also added.

Before

Re-submitted paper

2.3 Modeling of LNOP potential areas

…

This multidisciplinary approach has proven essential for effective management and prevention of illegal waste dumping and abandonment in the environment.

2.5 Software-based support

…

Discussion section

…

Findings clearly show that illegal waste disposal is not randomly distributed but is spatially and functionally conditioned. The most significant factors include the presence of degraded areas, distance from waste collection centers, vegetation density, terrain morphology, and degree of accessibility. These results are consistent with previous studies, which emphasize the importance of combining physical, social, and infrastructural variables.

2.3 Modeling of LNOP potential areas

…

These models are grounded in machine learning principles, where each new data entry contributes to improve the accuracy of predictions. The development of a methodology based on the concept of a circular data feedback loop enables continuous updating of the LNOP database and iterative adaptation of the risk area model. Spatial data from various influential layers, high-resolution orthophoto raster imagery, and LNOP data collected through a purpose-developed application environment serve as input datasets for cartographic modelling of risk zones. The entire workflow is highly scalable, as it allows for the flexible inclusion or exclusion of specific input datasets (variables), thereby enabling the development of alternative model scenarios with different input combinations for different land characteristics usable anywhere.

This multidisciplinary approach has proven essential for effective management and prevention of illegal waste dumping and abandonment in the environment.

2.5 Software-based support

…

The development of EkoVaruh establishes an efficient digital system that supports both operational fieldwork and systemic management of LNOP. At the same time, it encourages the active involvement of users and stakeholders in maintaining a clean environment. Compared to other application-based solutions—such as TrashOut [64], which also target illegal waste disposal—EkoVaruh is tailored specifically to the needs of local and national stakeholders and administrative systems. The EkoVaruh application is still under development and, as such, can be adapted to the needs of this type of mass data collection. In terms of its purpose, the application is being developed to be as simple and intuitive to use as possible, while still allowing administrators in the background to perform numerous analyses and queries. The potential for further development and enhancement is therefore undisputed.

Discussion section

…

· settlement patterns and house number density,

· public lighting infrastructure and power corridors,

· functionally degraded areas and land cover classes,

· LIDAR-based terrain features and vegetation masks,

· road and rail infrastructure.

This diversity enables a more comprehensive spatial analysis compared to previous studies that relied on a narrow set of predictors (e.g., population density, road proximity, landfill distance) (e.g., [40,65]). Our approach considers topographic data from LIDAR (elevation models), the presence of rail infrastructure, the electricity grid, functionally degraded areas, and the density of public lighting, allowing for a more holistic understanding of spatial dynamics. The approach is thus more robust and scalable.

The reduction in the surface area of high-risk zones—despite the increase in detected dumping sites—suggests that the predictive model is improving in spatial accuracy. Between Loop 1 and Loop 3, the number of recorded sites tripled, and waste volume more than doubled (+116%), affirming the method’s capacity for cumulative improvement. UAV surveys and targeted field inspections further validated the iterative enhancement of the model.

The approach offers strong potential for public sector implementation:

· Targeted monitoring: Municipalities and inspection bodies can use risk maps to optimize UAV inspections and enforcement operations.

· Digital civic engagement: The EkoVaruh application enhances public participation, aligning with principles of participatory environmental governance (e.g., LIFE Restart).

· Scalability and transferability: The model’s reliance on open-source tools and publicly available geospatial datasets makes it applicable across other Slovenian regions and transnational contexts (e.g., Northern Italy, Croatia, Austria).

Furthermore, the model could support the development of a regional environmental risk index to guide strategic planning, investment in remediation, and prioritization of enforcement efforts.

…

However, the modelling process is not without limitations. First, the input data is incomplete, as many LNOP sites remain unidentified—particularly in hard-to-reach areas or regions lacking active monitoring. Furthermore, data collection methods may introduce bias (e.g., self-reporting by citizens or targeted inspections). In addition, estimating the quantity and type of waste remains a challenge, as some attributes are based on subjective assessments. Most phases of the modelling process were limited to data from the MOM, meaning the approach is geographically constrained. The selected features and environmental variables were specifically tailored to the characteristics of this region; in other geographical contexts—such as coastal areas—different factors may be more relevant [40] and would require model adaptation.

…

It is worth highlighting that the study area considered in this research is relatively small compared to other available studies, while the number of identified LNOP is exceptionally high [66–69]. This provides a solid foundation for the development of a highly detailed spatial control model.

Another important distinction from other studies lies in the basic assumptions regarding the size of individual LNOP. In the area of MOM only a few large-scale sites were identified, with the majority being small and best described as “micro” LNOP. In contrast, other studies focus on significantly larger units for example, Rosa Jordá-Borrell et al. [69] define a lower size threshold of 2000 m² per site [69], while Nissim Seror and Boris A. Portnov [68] analyze LNOP ranging from 6 tons to 13,600 tons (with an average of 1,544 tons).

From a model performance perspective, it is also worth referencing the work of Lorenzo Carlos Quesada-Ruiz et al., who, similarly to this study, report a predictive accuracy exceeding 80% [66]. These findings support the conclusion that the present study is highly detailed and represents significant progress in terms of precision and methodological rigor.

…

Added references:

Glanville, K.; Chang, H.C. Mapping Illegal Domestic Waste Disposal Potential to Support Waste Management Efforts in Queensland, Australia. International Journal of Geographical Information Science 2015, 29, 1042–1058, doi:10.1080/13658816.2015.1008002.

Quesada-Ruiz, L.C.; Rodriguez-Galiano, V.; Jordá-Borrell, R. Characterization and Mapping of Illegal Landfill Potential Occurrence in the Canary Islands. Waste Management 2019, 85, 506–518, doi:10.1016/j.wasman.2019.01.015.

Tasaki, T.; Kawahata, T.; Osako, M.; Matsui, Y.; Takagishi, S.; Morita, A.; Akishima, S. A GIS-Based Zoning of Illegal Dumping Potential for Efficient Surveillance. Waste Management 2007, 27, 256–267, doi:10.1016/j.wasman.2006.01.018.

Seror, N.; Portnov, B.A. Identifying Areas under Potential Risk of Illegal Construction and Demolition Waste Dumping Using GIS Tools. Waste Management 2018, 75, 22–29, doi:10.1016/j.wasman.2018.01.027.

Jordá-Borrell, R.; Ruiz-Rodríguez, F.; Lucendo-Monedero, Á.L. Factor Analysis and Geographic Information System for Determining Probability Areas of Presence of Illegal Landfills. Ecol Indic 2014, 37, 151–160, doi:10.1016/j.ecolind.2013.10.001.

MINOR CONCERNS:

Role of UAV Technology: The use of UAVs for identifying new LNOPs is a modern and effective technique, as demonstrated. However, the role of the UAV should be clarified: is it primarily a validation tool or a discovery tool? A more nuanced discussion on the strategic deployment of UAVs would add significant practical value.

Author comment: The use of UAVs is described in detail in section 2.4 Remote sensing for LNOP recognition. A sentence has been added to clarify that drones were used in a targeted manner after applying a predictive model to verify whether there was any waste at the predicted locations.

Before

Re-submitted paper

As part of the assessment of LNOP, the applicability of remote sensing methods was also tested—specifically, the generation of high-resolution orthophoto raster imagery using UAV. The aim of this approach was to evaluate the potential for visual identification of additional LNOP and to assess the effectiveness of the method as a supporting tool for monitoring and documentation. Importantly, UAV deployment was guided by the outputs of the predictive LNOP risk model, serving as a targeted discovery tool to investigate areas with an elevated probability of illegal dumping.

Clarity of Figures and Captions: While the figures are generally informative, some could be improved: (1) The caption for Figure 13 contains a typo: "Figure 13Figure 8 displays...". This should be corrected. (2) The caption for Figure 16, the most critical result, could be more descriptive. It should explicitly state the key takeaway. (3) The maps (e.g., Figures 7, 9, 11) could benefit from a clearer legend or context, especially for readers unfamiliar with the geography of Maribor.

Author comment: The caption of Figure 13 is corrected. The caption for Figure 16 (now Figure 17) is now more descriptive. The maps in Figures 7, 9 and 11 are now clearer (bigger legend and scale).

Other changes to the paper:

Author comment: Some methodological limitations were already indicated in the first version of the paper (in the Discussion section). However, we have added a few more critical points and limitations, and suggested potential improvements to the model for future studies.

Before

Re-submitted paper

Discussion section

…

Conclusion section

· …

Discussion section

…

The approach offers strong potential for public sector implementation:

· Targeted monitoring: Municipalities and inspection bodies can use risk maps to optimize UAV inspections and enforcement operations.

· Digital civic engagement: The EkoVaruh application enhances public participation, aligning with principles of participatory environmental governance (e.g., LIFE Restart).

Furthermore, the model could support the development of a regional environmental risk index to guide strategic planning, investment in remediation, and prioritization of enforcement efforts.

Validation of the prediction model also presents a significant challenge due to the nature of the available data and the algorithms used. Since the dataset follows a positive-unlabeled (PU) structure—where only known illegal dumping sites are labeled and all other locations remain unlabeled—the absence of confirmed negative examples makes it difficult to objectively evaluate model accuracy using traditional validation methods (e.g., precision, recall, ROC curves). The algorithms employed (One-Class SVM, Isolation Forest, and the Elkan-Noto method) are specifically designed for scenarios where true negatives are unknown, but this also limits the possibility of robust cross-validation or ground truth comparison. Consequently, any assessment of model performance must rely on indirect evidence, expert interpretation, or future detection of new LNOP cases in predicted high-risk areas, which can then serve as post hoc validation. This methodological constraint is inherent to many real-world anomaly detection problems and remains a subject of ongoing research.

Conclusion section

· …

For long-term effectiveness, a systemic approach at the national level will be essential. This includes the standardization of data structures, mandatory reporting obligations, and harmonized implementation of circular data feedback loops across all local communities. Only such an approach can ensure data consistency, result comparability, and the establishment of a robust national LNOP monitoring system. Also, public sector engagement, targeted monitoring, visual inspections, EkoVaruh application improvement, application across other Slovenian regions and over the state border and more consistent citizen participation can be improved and intensified.

Author comment: The policy implications of the results obtained are now clearer with some more explanation in Discussion and Conclusion section.

Before

Re-submitted paper

Conclusion section:

…

Discussion section:

…

Conclusion section:

…

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper presents a risk model using a circular data feedback loop to predict/detect illegal dumping sites in Maribor, Slovenia based on machine learning. By integrating field data, UAV surveys, and GIS modeling, the proposed methodology improves detection of smaller, hard-to-spot dumping events and supports targeted, technology-driven waste management. The paper has significant research value, but some major issues need to be addressed before it can be accepted.

1- The introduction should be shortened to focus only on the core problem of illegal dumping, its environmental and economic impacts, and the study’s main objective. Irrelevant global case studies, literature reviews, and technical comparisons should be removed. However, Slovenia-specific context and legislative references should be retained to support the case study. A new "Related Work" section should be added after the introduction. This section should organize all previously scattered literature, including global dumping drivers, detection technologies, risk factors, and limitations of existing methods. Tables and numerous references from the original introduction should be moved to this section to streamline the structure while clearly distinguishing this study’s innovations in PU learning and the EkoVaruh app.

2- In-text citation errors (Page 23-line 947): Error! Reference source not found."

3- Revise all references (Standardize formatting, add missing URLs/DOIs, and correct author names). In fact, some references use bold text, while others do not, missing DOIs/URLs, incomplete author lists, missing date,…

4- Contents of Figure 1 are unclear.

5- The paper is well-written with only minor grammatical/formatting issues:

Correct article omissions (*a*/an/the), ensure consistent hyphenation in compound terms (e.g., semi-supervised), fix placeholder text (e.g., "Error! Reference source not found"), refine long sentences for conciseness, verify parallel structure in lists.

These edits are minimal and do not detract from the paper's substantive quality. The manuscript is publishable as-is but would benefit from a final proofread to polish these minor details.

6- Quantify uncertainty in PU learning (e.g., via bootstrapping) and compare model performance against baseline methods (e.g., logistic regression). Ground-truth "negative" samples can be also included to improve statistical rigor.

7- If possible, validate the model in contrasting Slovenian regions or cross-border to test transferability.

8- Integrate UAV photogrammetry/LIDAR for 3D waste volume estimation to reduce subjectivity. Test multispectral sensors for waste-type classification.

9- Include socioeconomic variables (e.g., income, education) from census data into the risk model to enhance predictive power.

10- Compare costs of UAV-led monitoring vs. traditional inspections to highlight economic efficiency for policymakers.

11- Prioritize AI development (e.g., CNN for orthophoto analysis) to reduce manual validation. Clarify timelines for EkoVaruh AI upgrades.

12- Define "micro-LNOP" (Section 4) with size thresholds to distinguish from larger illegal landfills in comparative studies.

Comments on the Quality of English Language

The paper is well-written with only minor grammatical/formatting issues:

These edits are minimal and do not detract from the paper's substantive quality. The manuscript is publishable as-is but would benefit from a final proofread to polish these minor details.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

"Modelling of risk areas with potential for illegal dumping: integrating databases, GIS analysis and remote sensing into a circular data loop" includes an important environmental and societal issue. The development of a GIS spatial-modeling-based machine learning approach along with remote sensing on UAV and a circular data feedback loop underscores monitoring from illegal dumping and is very relevant and well within the scope of Sustainability. The paper pursued a strong methodological approach mixing multivariate data sources and technology for developing a predictive model for illuminating illegal waste dumping places: this, interdisciplinary, is of novel thought and directly applicable in practice for waste-management authorities.

The manuscript provides a thorough coverage and is well organized; however, the clarity of certain figures, captions, and text elements needs to be enhanced. More conciseness in some parts (mainly the introduction) would increase readability, although the methods and results are thus detailed. Language editing is also recommended to make sentences less complex and the writing more fluid.

Figures (e.g., Figures 6–15) are crucial for understanding the results, but some are low resolution and lack sufficient labeling. Captions should be more descriptive, allowing the figure to be interpreted without fully referring to the text.
While the manuscript is generally well written, some sentences are long and complex, making them difficult to follow. For example, sections of the introduction and methodology contain extended sentences that could be broken down for clarity. A professional language editing service or thorough proofreading is recommended.
The introduction provides an extensive overview of the problem and related research. While informative, it could be shortened to improve focus on the main research gap and the novelty of the proposed approach.
Although the manuscript mentions the circular data feedback loop, the authors should more explicitly highlight how this approach differs from previous GIS and UAV-based monitoring methods. State the specific added value of combining PU learning with circular data loops.
The paper would benefit from a short paragraph discussing the main limitations of the proposed approach (e.g., dependency on UAV flight permissions, seasonal vegetation influence on detection, and resource requirements for large-scale application). Additionally, discuss the scalability of the methodology to other regions or countries.
The development and potential of the EkoVaruh application are interesting, but more details on its current functionalities and integration capabilities with national systems could strengthen the discussion.
Include a clearer discussion on how municipal authorities or policymakers can implement this approach, considering budget and operational constraints.
The references are generally appropriate and relevant. However, ensure that all sources cited in the text (e.g., figures, regulations) are included in the reference list and formatted according to journal requirements.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author responded constructively to the previous round of review, but there are still areas for improvement.The suggestions are as follows：

Although accuracy is a useful indicator, it can be misleading for unbalanced data sets. To further strengthen verification, indicators such as precision, recall and F1 scores are needed, and the ability to distinguish between high-risk and low-risk areas without being distorted by a large number of negative samples.
Further discussion is needed on the limitations of PU learning methods. Specifically, As the core method of the paper, can the author provide a detailed explanation of the strategy used to sample "unlabeled" data points from the study area?
The challenges of data reconciliation and the minimum data requirements for model validity need to be discussed, for example, the model integrates many specific GIS datasets. To what extent are these data sets standardized in Slovenia or other EU countries?
‘circular data loop’ is a central concept in this study, and the authors can provide a more explicit explanation of this improved mechanism in the “Discussion” section, which provides a better understanding of the mechanism of data looping, how exactly does the new data from each iteration improve the model's understanding of risk factors?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Done

Author Response

We would like to humbly and gratefully thank you for your professional, constructive review of the article.