Balancing wetland conservation under disease risk in Indonesia: A spatial MCDA approach

Zhu, Yiting; de Ruiter, Marleen; Buijs, Sophie; van Maanen, Nicole

doi:10.5194/egusphere-2026-94

Preprints

https://doi.org/10.5194/egusphere-2026-94

Preprints

15 Jan 2026

| 15 Jan 2026

Balancing wetland conservation under disease risk in Indonesia: A spatial MCDA approach

Yiting Zhu, Marleen de Ruiter, Sophie Buijs, and Nicole van Maanen

Abstract. Wetlands provide essential ecosystem services but can also serve as breeding habitats for disease vectors such as mosquitoes, creating complex challenges for conservation planning. Indonesia has extensive wetlands and high malaria incidence, requiring conservation strategies that integrate both ecological and health considerations. This study implements a spatial Multi-Criteria Decision Analysis (MCDA) framework to support wetland conservation planning by integrating ecological benefits and vector-borne disease risk. The analysis integrated eight criteria using literature-informed weighting across 94.6 % of Indonesia's wetland areas. Results reveal that conservation and health factors operate largely independently (r = 0.099, p < 0.001), suggesting minimal trade-offs between objectives. The findings demonstrate that wetland conservation and health objectives are compatible in most regions, enabling strategies that optimize ecological outcomes without systematically increasing disease exposure. Papua is noted as a region of interest, being the main region where high ecological value does coincide with elevated disease risk. The framework supports conceptualizing wetlands as Nature-based Solutions that simultaneously deliver conservation and public health benefits, providing practical guidance for Indonesian policymakers and a replicable template for other tropical regions facing similar conservation-health challenges.

Received: 07 Jan 2026 – Discussion started: 15 Jan 2026

Competing interests: At least one of the (co-)authors is a member of the editorial board of Natural Hazards and Earth System Sciences.

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

Download & links

Yiting Zhu, Marleen de Ruiter, Sophie Buijs, and Nicole van Maanen

Status: final response (author comments only)

RC1: 'Comment on egusphere-2026-94', Anonymous Referee #1, 09 Mar 2026

The manuscript is titled “Balancing wetland conservation under disease risk in Indonesia: A spatial MCDA approach”. The authors present a spatial Multi-Criteria Decision analysis framework to support wetland conservation planning. The presented framework is applied to the case of Indonesia. The topic of spatial planning for Nature-based Solutions is an emerging and relevant research topic, especially given the increasing pressure by climate change and urban densification. This manuscript focusses on vector-borne disease risk assessment and potential trade-offs to ecological value, which is both within the scope of NHSS and a highly interesting case for spatial MCDA methods. The main contributions are tied to the issue of integrated wetlands management and practical insights for decision-makers in the examined region, as the data and MCDA methodology are sourced from literature.

General comments:

1) Overall, this is a mostly well-conducted piece of research, clearly outlining the methods being used, leading to a substantial set of conclusions that is widely supported by the results being presented. It is well written, of adequate length and easy to understand for an audience with preliminary knowledge in related fields.

2) However, a few methodological details need to be clarified and could be presented in more comprehensive manner. Furthermore, some assumptions may merit better reasoning or handling in the analysis. While the line of argumentation in the discussion is overall clear and well-reasoned, the methodological approach is not discussed at all. This could be extended.

Some more specific comments on this:

1) Page 2, l. 32: You write that “wetlands can also help reduce disease risk when well-managed, [..]”. To me, this implies a potential trade-off between management effort and disease risk that should be incorporated into the system of criteria. Currently this dimension is not reflected at all in the assessment. If not represented by actual criteria when evaluation priorities for intervention, at least, it should be discussed later when talking about real-world implications of the conservation priority map.

2) Page 4, Table 1: I am wondering about the criteria selection process. Why were specifically these criteria selected? It is mentioned that they are “established criteria from literature review”. For replicability in slightly different cases, the actual process of narrowing down a literature review towards a selection of final criteria could be interesting.

3) Page 6, ll. 122: The normalization process, literature sources and assumptions are well-documented. However, as there are assumptions being taken, I suggest that the sensitivity of the results towards changes in these assumptions on the results should be examined in sensitivity analyses as often recommended for MCDA.

4) Page 6, l 138: You mention the term “compensatory elements”. What do you mean by that? To me this remains unclear throughout the paper, especially related to the property of compensation in MCDA approaches. What exactly is non-compensatory and what is the exact meaning for the MCDA aggregation?

5) Page 6, ll. 142: I may be wrong, but to me the weights don’t add up. The sum of hazardous factors is 50% according to my calculation approach (25% malaria, 15% elevation, 6% temperature, 4% precipitation).

6) Page 6, ll. 142: How exactly were weights derived from the literature? What was the process of condensing information in the literature to a set of weights? To me, this is currently not apparent.

7) Page 7, l. 172: “[…] we constructed a composite ecological benefit index by integrating biodiversity significance, carbon storage potential, and water provision capacity.” How exactly was this done? How were criteria states on these three criteria that you mention translated to the dimensionless ecological benefit index?

8) Page 8, Subsection 2.5: At this point, you might consider adding a comprehensive table, summarizing all inputs to the MCDA (criteria, S_i, w_i) for the standard analysis and the scenarios.

9) Page 8, Subsection 2.5: I find the way the scenarios are set up problematic as there is a mismatch between criteria. As the criteria set is currently set up, there is only one ecological composite criterion while there are 8 disease risk criteria. This biases the overall priority score, as defined in eq. (2) towards ecological benefits, as only one criterion must reach a score of 1 to obtain 1 ecological priority. To obtain 1 for disease risk, all 8 criteria must reach the worst state. This makes comparability of the scenarios difficult. As you can see in Figure 3, the priority index is heavily biased by the weight applied to ecological priority. To me, the scenarios require a redesign by (i) either accounting for the bias in the distribution of lower-level domain weights (individual criteria) instead of just altering higher-level domain weights (ecological weight/risk weight) and/or (ii) examining further sources of uncertainty, e.g., uncertainty in input data.

10) Page 11, Subsection 3.3: To me the inversion from conservation priority rankings to benefit scores is not completely evident. You state that areas with intact ecosystems receive the highest benefit score (approaching 1), while urgent conservation intervention areas receive lower scores (approaching 0). Here, a lower score corresponds to higher priority. For the risk indicators, it appears opposite to me. The higher the risk, the higher the score and thus the priority. I suggest to either elaborate or clarify this in the manuscript.

11) Page 13, ll. 9–10: “Despite the different weighting emphases, all three scenarios identify similar geographic regions as important, indicating robust identification of key conservation areas. Following my previous comment, this does not emerge from the previously stated information. Shouldn’t areas with currently low ecological benefits receive high priority as these areas with high disease risk? Visually observing the ecological benefit map in Figure B4, the lower left promontory of Papua is shaded in relatively dark green color, indicating high conservation benefit and thus low priority. Under very high weighting of ecological benefits, the overall conservation priority of this area should be somewhat lower, right? I suggest to clarify the purpose of the scenario comparison and further discuss the insights.

Additionally, please note the previous remark on the scope of the sensitivity analysis, which may prohibit encompassing robustness assessments.

12) Page 15, ll. 387–388: As you mention “participatory planning processes that incorporate local stakeholder priorities”, it would be interesting to researchers and practitioners which are seeking to apply the proposed framework to discuss how exactly the framework could be made participatory and how exactly stakeholder priorities could be elicited and included.” I suggest to at least briefly discuss it. Does this only refer to criteria weights or also to more spatial information that is elicited from stakeholders? Maybe even include it in further research as the elicitation of spatial preferences is currently only briefly explored in MCDA literature.

Citation: https://doi.org/10.5194/egusphere-2026-94-RC1

AC1: 'Reply on RC1', Yiting Zhu, 19 Jun 2026

We sincerely thank Reviewer 1 for taking the time to review our manuscript. We deeply appreciate your positive overall assessment of our research and your highly constructive suggestions. Your insights have significantly helped us improve the clarity, transparency, and practical value of our work. Please find our point-by-point responses below.

We sincerely thank the reviewer for this insightful observation. We completely agree that wetland management efforts (e.g., maintaining water flow, protecting mosquito-predator habitats) play a crucial role in mitigating disease risk. However, as the reviewer correctly anticipated, "management effort" is highly localized and extremely difficult to quantify as a continuous spatial variable at a national scale for an MCDA model. Instead of mapping management directly, our framework indirectly captures the result of good management through the Ecological Benefit Index, where ecologically intact and functional wetlands receive higher scores.

This dimension is already touched on in two places in the current manuscript. In the Introduction, we discuss the mechanisms by which management reduces risk (restoring hydrological flow, supporting predator species, optimizing land-use zoning), as well as the converse pathway about how degradation and unmanaged land-use change increase vector-human contact. In Section 4.2, we propose a tiered resource-allocation strategy and note that "targeted and grounded work must be undertaken by local governments and communities" to address local conditions our national-scale analysis cannot capture.

To respond more directly to the reviewer's point, we propose to revise Section 4.4 to explicitly connect this limitation to our discussion of participatory planning. We will add a sentence clarifying that local management practices and priorities, not only criteria weights are a natural target for stakeholder elicitation, but locally-sourced data on management intensity could be incorporated as an additional criterion or local adjustment layer once finer-resolution data become available.

We appreciate the reviewer's suggestion to improve the transparency and replicability of our criteria selection process. Our selection was systematically narrowed down by reviewing previous spatial MCDA studies focusing on malaria and vector-borne disease risk mapping (e.g., studies in Madagascar, Ethiopia, and China-Myanmar borders). We do have a table listing all the literature, but we didn’t include it in the manuscript. Please find the detailed table below. This table details the specific studies we reviewed, the MCDA models they utilized, their selected key criteria, and how they determined their weights.

Study Region & Reference	MCDA Model Used	Key Criteria & Weights	How Weights Were Determined
Yunnan, China (China–Myanmar border) (Zhao et al., 2020)	GIS-based Spatial MCDA; AHP for weighting	elevation (29.7%), imported cases (25.8%), distance to a water body (17.5%), cultivated land (8.6%), human population density (4.3%), forest coverage (4.1%), urbanization (3.6%), distance to a health facility (3.3%), and distance to road (3.0%)	Expert elicitation (6 experts), AHP pairwise comparisons, Consistency Ratio (CR)<0.1
Madagascar (Central highlands) (Rakotoarison et al., 2020)	GIS-based MCDA Spatial for risk mapping; AHP for weighting	Population density (49.9%), Distance to wetland (18.2%), Temperature (17.0%), Elevation (9.1%), Precipitation (5.8%)	Expert elicitation (7 experts), AHP pairwise comparisons, CR=0.06, sensitivity analysis
Northern South America (Inland) (Alimi et al., 2016)	GIS-based Spatial MCDA for risk mapping; AHP for weighting	Distance from deforested patches (9.96%), Population density (5.93%), Distance from roads (3.79%), Distance from urban areas (4.2%), Distance from wetlands (13.91%), Elevation(16.8%), Precipitation (17.84%), Temperature (20.06%), Topographic Wetness Index (7.51%)	Expert elicitation, AHP pairwise comparisons
South West Ethiopia (Didessa District) (Gebre et al., 2020)	GIS-based Spatial MCDA for risk mapping; AHP for weighting	Malaria Hazard Factors (0.6): altitude (18.07%), rainfall (38.32%), slope (14.96%), temperature (22.78%), river (5.87%) Elements at risk factors(0.2): Land use land cover change (60%), population density (40%) Vulnerability factors (0.2): distance to health institutions	AHP pairwise comparisons by authors, CR<0.1
Ghana (Inland) (Kumi-Boateng et al., 2015)	GIS-based Spatial MCDA for risk mapping; AHP for weighting	Rainfall (23%), Temperature (22%), Potential Evapo-Transpiration (14%), Distance to Road (4%), Distance to Water Bodies (12%), Land Cover (6%), Slope (5%), Altitude (14%)	AHP pairwise comparisons by authors, CR = 0.057

We sincerely appreciate this methodological suggestion and agree that systematically testing the sensitivity of MCDA results to underlying assumptions is recommended practice. We would like to address this in two parts, corresponding to the two main sources of assumption in our model: the criteria weights, and the normalization functions.

For the weights, our existing scenario comparison already provides a macro-level sensitivity test of how the conservation priority ranking responds to shifts in domain-level weighting. To strengthen this further, we propose to add an additional equal-weighting scenario, in which all individual criteria receive equal weight, as a neutral baseline against which the literature-derived weighting scheme can be compared. This will help clarify the extent to which our results are driven by the specific weights chosen versus the underlying data structure itself.

For the normalization functions, a full parameter-by-parameter sensitivity analysis would require reprocessing the full national raster (21.9 million pixels) under many parameter combinations, which is computationally intensive and beyond the scope of this study's national-scale, policy-oriented aim.
We apologize if the explanation in the manuscript was not explicit enough. We defined these concepts in Section 2.4 (Lines 135-140).

To clarify:

- Compensatory aggregation means using a weighted linear combination where a high score in one risk factor can "compensate" or offset a low score in another. For instance, high mountain areas have low environmental suitability (low score), but if they have high population density and land-use change (high score), the overall risk score remains elevated.

- Non-compensatory elements act as "strict thresholds". In our study, this refers to the spatial masking of wetlands. If a pixel is not a wetland, it receives an absolute zero for conservation priority, regardless of how high its disease risk or other ecological scores might be. No other high scores can "compensate" for the fact that it is not a wetland.
We sincerely apologize for this typographical error and are very grateful to the reviewer for catching it. The reviewer's calculation is absolutely correct. The hazard factors indeed sum up to 50% (25% + 15% + 6% + 4%), with exposure factors at 25% and vulnerability factors at 25%, making the total 100%.
We thank the reviewer for requesting this important clarification. The weights were not derived through a strict mathematical averaging of previous literature, but rather through a structured "expert adaptation" process based on the relative importance established in previous MCDA studies, tailored to our specific hazard-exposure-vulnerability framework for Indonesia.

Specifically, the process involved two main logical steps:

(1). Relative Ranking from Literature: We reviewed previous spatial MCDA malaria studies (now detailed in the newly added table). These studies consistently show that elevation factors generally carry more weight than climatic factors in tropical environments. We maintained this relative hierarchy in our hazard weights (Elevation 15% > Temperature 6% > Precipitation 4%).

(2). Contextual Adaptation for Indonesia: Unlike many purely environmental risk models in the literature, our study had access to highly detailed spatial data on actual malaria incidence. Therefore, we purposefully assigned the highest individual weight (25%) to this empirical epidemiological evidence, adjusting the remaining environmental and socio-economic weights proportionally around it to fit the 50% (Hazard) - 25% (Exposure) - 25% (Vulnerability) structural design.

We acknowledge in the Limitations section that while these literature-informed weights serve to demonstrate the utility of our national-scale framework, future localized implementations should elicit context-specific weights directly from local stakeholders (e.g., using AHP methodologies).
We thank the reviewer for pointing out this lack of clarity in our description. We would like to clarify that the integration of biodiversity, carbon, and water services was not performed by the authors of this study. Instead, we utilized a pre-existing, peer-reviewed composite index from the "Areas of global importance for conserving terrestrial biodiversity, carbon, and water" dataset (Jung et al., 2021). To translate this dataset into a dimensionless "Ecological Benefit Index" suitable for our MCDA, we performed an inverse normalization of the original priority rankings as stated in Appendix Table A1. We accept the text on Page 7 to replace "we constructed" with "we utilized the composite index from Jung et al. (2021)" to accurately reflect the source of this data.
We thank the reviewer for this suggestion to improve the readability and transparency of our methodology. As we response at point 3, we propose to add an additional equal-weighting scenario for the sensitivity analysis to further improve validation and comparison of the methodology.
We sincerely thank the reviewer for this important critique, which raises two separable issues: the mathematical structure of the aggregation, and the empirical reason the three scenario maps appear spatially similar despite different weights. We address each in turn, and propose an additional analysis to investigate the underlying concern more directly.

(1). The aggregation is a balanced 1 vs 1 structure, not 1 vs 8:

Equation (2) does not aggregate 1 ecological criterion against 8 individual risk criteria. The 8 risk criteria are first internally aggregated to produce a single Disease Risk Index (0-1). Equation (2) then aggregates this Risk Index with the composite Ecological Benefit Index (0-1). Therefore, the scenario weights are applied fairly to two dimensionless domain scores (1 vs 1). A pixel only needs the weighted sum of its risk factors to be high to achieve a high Risk Index; it does not need all 8 criteria to be at their worst state.

That said, we recognize that the reviewer's underlying concern remains valid even within this hierarchical structure: because the Risk Index is itself a weighted average across eight criteria, it is statistically less likely to reach extreme values than the Ecological Benefit Index, which is built from fewer underlying components. To investigate how much this contributes to the pattern in Figure 3, separately from the domain-level scenario weights, we propose to add a sensitivity analysis in which the eight disease-risk criteria are assigned equal weight as replied in point 3.

(2). Why the scenarios show similar spatial trends:

As the standardized maps show, regions such as Papua exhibit high values on both the Ecological Benefit Index and the Disease Risk Index simultaneously. Where both domain scores are high, any combination of scenario weights will mathematically yield a high overall priority score for that pixel. The similarity across scenarios therefore reflects this empirical co-occurrence in the underlying data, rather than a weighting artifact. The scenarios primarily shift the relative priority ranking of moderate-risk, moderate-benefit areas (e.g., parts of Sumatra and Kalimantan) rather than the overall geographic pattern.
We thank the reviewer for pointing out this confusing phrasing. To clarify our logic: our framework focuses on the protection of highly intact ecosystems rather than the restoration of degraded ones. Because we strictly measure current ecological benefits, intact ecosystems receive a high benefit score (approaching 1.0), while degraded areas receive a lower score (approaching 0). For example, heavily modified areas like Java have much lower current ecological benefits compared to the dense, intact forest in Papua.

Mathematically, our MCDA formula simply adds “ecological benefit” and “disease risk”. An area with a highly intact ecosystem and a high disease risk receives the maximum overall priority. This correctly identifies regions like Papua as top priorities for integrated planning. Therefore, reversing the original dataset into a “benefit score” was a necessary step.
We thank the reviewer for this comment. The first part of this question stems from the exact same phrasing confusion addressed in our response to Comment 10.

To reiterate, our framework prioritizes the protection of intact ecosystems, not the restoration of degraded ones. Therefore, areas with high ecological benefits (the dark green areas in Figure B4, such as Papua) receive a Benefit Score approaching 1.0. Because our formula uses additive benefits, applying a very high weight (70%) to the ecological component means that Papua's high Benefit Score (1.0) mathematically drives its final conservation priority score higher, not lower. This is why Papua consistently emerges as a top priority across all scenarios, correctly reflecting its dual status as a highly intact ecosystem and a high disease-risk zone.

We also agree with the reviewer's critique of the word "robust" in this context. In light of our response to point 3 and 9, with acknowledging the limited scope of our sensitivity analysis, we propose an additional equal-weighting analyse.
We thank the reviewer for this suggestion. You’ve touched on exactly why we believe this framework is useful for the future. As we noted in our limitations, the weights in this study are informed by literature to establish a national-scale baseline. However, this is just the starting point. When we talk about participatory planning, we mean that local practitioners can take this modular framework and determine their own priorities.
For example, stakeholders in a specific region could use AHP to adjust the criteria weights to better reflect their local challenges. Furthermore, they could provide spatial preferences through participatory GIS, where local people actually mark out high-risk spots or protected zones on the map that satellite data might miss.

Citation: https://doi.org/10.5194/egusphere-2026-94-AC1

RC2: 'Comment on egusphere-2026-94', Anonymous Referee #2, 25 Apr 2026

This study presents an author-led desktop study GIS-MCDA of vector-borne disease risk and wetland conservation prioritisation. While such study can be of great value to planners and policy-makers, I have concerns regarding the methodological approach and the academic contribution provided by the article. In the text below I substantiate these concerns step by step, following the structure of the paper.

First of all, there is limited engagement with the field of MCDA and the philosophy within this field that was taken. That also shows in the language used. I would advise the authors to engage with the different approaches that the broad decision science, operational research field encompasses, and their position in it. Selecting a specific philosophy (such as multi-attribute value theory MAVT) and following its analytical framework more strictly could solidify the methodology and make it more aligned with academic standards while improving its reproducibility. Several questions emerge from this lack of engagement with theory, a few of which I list here:

How did the literature review lead to the list of criteria and its categorisation (building what in MAVT we call an objectives hierarchy)? And is this process reproducible? Where can the reader find the steps that lead to the list?
With the value scaling did you use absolute or relative scales (in other words, is there some “objective” standard for each of the criteria (but that means that the variation in your dataset might be minimal for some criteria if you data is relatively homogenous), which could be globally applied, or does it depend on the local data and is thus relative (to enable distinction within your specific case study? And why did you apply that? What are the consequences for the interpretation of your findings?
The weighting procedure is untransparent and seems rather arbitrary, while it has decisive influence on the outcome of the study (just as the decision to use full compensation, which should also be the result of stakeholder preference elicitation). As described currently the step is not reproducible, making the results of the study speculative and pivoting on this step to a unsatisfactory degree.
The absence of a sensitivity analysis where the effects of varying weights, but also the value scales and degree of compensation are explored (the latter often being the most impactful of all) is particularly problematic in this context (the sensitivity analysis described in line 190 is not a true, systematic sensitivity analysis following MCDA methodological standards).

Furthermore, here are my remaining questions, suggestions and comments on specific parts of the paper:

Line 165: what was the resolution of each of the datasets, and did you need to do a lot of interpolation in this step? What are the consequences, and why was this resolution selected? What are the consequences for interpretation (e.g. do mosquito’s travel over such distances?)
Line 192: I believe that this may be one of the most interesting contributions of the paper. However, there are no details on the methodology, making it hard to understand how this was done.
Lines 265 – 270: An MCDA just outputs what you put in. However, the factors that you are describing here to have led to the high risk visible on the map in Papua references factors that are not part of the MCDA, at least not directly.
285-288: this seems to me the main contribution of the paper. The other results are merely artifacts of the decisions made by the authors and a spatial representation and combination of data, rather than a true GIS-MCDA that combines stakeholder preferences with real-world spatial data representing suitability for certain interventions.
I don’t understand how the authors arrive at the conclusion stated in line 314
While the conclusions in 320 seem useful to practice, I wonder if we couldn’t have arrived at the same conclusions without doing the exercise, considering it was merely a combination of information rather than a true analysis of interactions or the combination of subjective with objective knowledge. This puts the academic value of the analysis into question, and I invite the authors to reflect on the academic contribution: what knowledge gap is being addressed and why is this novel?
I wonder what exactly the authors mean with “framework” that can be replicated in other places in the world? If it is the collection of datasets and the value scales and weights for the MCDA, I would argue this is not a proven and tested set of MCDA settings that can be considered a generalisable “framework”, as no empirical testing of this framework is described in the paper, and thus its validity remains uncertain.
Line 423: The conclusion suggest there was a breakthrough methodologically speaking, but what the breakthrough is remains unclear. Standard MCDA methods were applied with mostly open datasets and author-assigned parameters.

Finally, please find some textual comments in the table below:

Line	Comment
64	NbS are not a framework, but a solution.
72	Would be good to refer to this massive body of literature by giving some practical examples of comparable studies done in the past (there are thousands) and cite a textbook, like Malczewski & Rinner (2015).
Figure 1	This figure is missing a legend/caption explaining what the different colours indicate. Also, I don’t understand why phase 2 and 3 are separated, as the components of phase 3 are part of spatial MCDA.
123	“standardization” in GIS-MCDA we call this “value scaling”. This is typically a process done with knowledgeable stakeholders. “normalization functions” are called “value functions” (in MAVT) or “utility functions” (in MAUT).
133	Minimum, or the maximum performance among criteria counts in non-compensatory methods. Then, there is a whole world in between of partial compensation.
Equations	Please use mathematical conventions to create the equations, e.g. using w with subscript for the different types of weights. Please use MCDA textbooks to familiarise yourself with these conventions.

Citation: https://doi.org/10.5194/egusphere-2026-94-RC2

AC2: 'Reply on RC2', Yiting Zhu, 19 Jun 2026

We sincerely thank the reviewer for this thorough and constructive evaluation of our manuscript. The comments have pushed us to engage much more carefully with the MCDA literature and to be far more transparent about our methodological choices, particularly around weighting, value scaling, and the scope of our sensitivity analysis. We believe the revisions prompted by this review meaningfully strengthen the rigor and clarity of the paper, and we address each point below.

We thank the reviewer for raising this point, and have compiled a focused review of comparable spatial MCDA studies applied to vector-borne disease risk mapping in tropical and sub-tropical settings: Yunnan, China (Zhao et al., 2020); Madagascar (Rakotoarison et al., 2020); Northern South America (Alimi et al., 2016); South West Ethiopia (Gebre et al., 2020); and Ghana (Kumi-Boateng et al., 2015). All five studies adopt the same general approach we follow here: a GIS-based, weighted linear combination of hazard, exposure, and (where included) vulnerability factors, integrated into a continuous risk surface. This confirms that our overall methodological framing is consistent with established practice within this specific body of disease-risk MCDA literature. Because of the strong foundation of the selecting literature and the rage of each criteria, the selection is definitly reproducible to other place as well. We acknowledge this as a methodological choice distinct from the AHP convention in this literature, but out reseach is meant to build a broad framework to analysis the relation between disease risk and ecological benifit.

Study Region & Reference	MCDA Model Used	Key Criteria & Weights	How Weights Were Determined
Yunnan, China (China–Myanmar border) (Zhao et al., 2020)	GIS-based Spatial MCDA; AHP for weighting	elevation (29.7%), imported cases (25.8%), distance to a water body (17.5%), cultivated land (8.6%), human population density (4.3%), forest coverage (4.1%), urbanization (3.6%), distance to a health facility (3.3%), and distance to road (3.0%)	Expert elicitation (6 experts), AHP pairwise comparisons, Consistency Ratio (CR)<0.1
Madagascar (Central highlands) (Rakotoarison et al., 2020)	GIS-based MCDA Spatial for risk mapping; AHP for weighting	Population density (49.9%), Distance to wetland (18.2%), Temperature (17.0%), Elevation (9.1%), Precipitation (5.8%)	Expert elicitation (7 experts), AHP pairwise comparisons, CR=0.06, sensitivity analysis
Northern South America (Inland) (Alimi et al., 2016)	GIS-based Spatial MCDA for risk mapping; AHP for weighting	Distance from deforested patches (9.96%), Population density (5.93%), Distance from roads (3.79%), Distance from urban areas (4.2%), Distance from wetlands (13.91%), Elevation(16.8%), Precipitation (17.84%), Temperature (20.06%), Topographic Wetness Index (7.51%)	Expert elicitation, AHP pairwise comparisons
South West Ethiopia (Didessa District) (Gebre et al., 2020)	GIS-based Spatial MCDA for risk mapping; AHP for weighting	Malaria Hazard Factors (0.6): altitude (18.07%), rainfall (38.32%), slope (14.96%), temperature (22.78%), river (5.87%) Elements at risk factors(0.2): Land use land cover change (60%), population density (40%) Vulnerability factors (0.2): distance to health institutions	AHP pairwise comparisons by authors, CR<0.1
Ghana (Inland) (Kumi-Boateng et al., 2015)	GIS-based Spatial MCDA for risk mapping; AHP for weighting	Rainfall (23%), Temperature (22%), Potential Evapo-Transpiration (14%), Distance to Road (4%), Distance to Water Bodies (12%), Land Cover (6%), Slope (5%), Altitude (14%)	AHP pairwise comparisons by authors, CR = 0.057

The newly added tale also clarifies our criteria selection process: criteria such as elevation, temperature, precipitation, population density, and distance to water bodies recur consistently across this comparator literature as established disease-risk factors, which informed the hazard and exposure components of our objectives hierarchy. We also admit that this study relies on globally available datasets (e.g., satellite-derived climate, population, and land-cover products) rather than locally collected, ground-truthed data. Thus, it might not be able to capture locally fine situations. We stated the limitation that should consider local stakeholder’s participation and adjust the scaling in future study, this study is more about to provide a national scale relationship of malaria risk and the wetland conservation.
As detailed in our response above, we have added new Table, documenting the literature sources underlying our criteria and weights. We acknowledge that, unlike the AHP-elicited weights used in the comparator studies, our weights were derived directly from literature evidence rather than independent expert elicitation, we discuss below how this could be strengthened in future applications. The selection of the threshold was based on the absolute data range provided by the reference literature, and it could be applied globally. We primarily utilized absolute scales based on objective biological and geographical thresholds established in literature, rather than relative scales derived only from our dataset's internal variation. For example, the value function for temperature is based on empirical entomological evidence indicating optimal transmission at 25-26°C.
We agree that the analysis described at Line 190 is a comparison of discrete domain-weighting scenarios rather than a systematic, parameter-level sensitivity analysis, and we will revise this language accordingly. To strengthen this further, we propose to add an additional equal-weighting scenario, in which all individual criteria receive equal weight, as a neutral baseline against which the literature-derived weighting scheme can be compared. This will help clarify the extent to which our results are driven by the specific weights chosen versus the underlying data structure itself. We acknowledge, however, that a full systematic sensitivity analysis, which separately varying each normalization threshold and testing alternative degrees of compensation, would require reprocessing the full national raster (21.9 million pixels) under many parameter combinations, which is computationally intensive and beyond the scope of this study's national-scale, policy-oriented aim.

Response to the specific comments:

Line 165: We thank for the reviewer’s detailed focus on the data analysis to improve the trasparancy. Native resolutions of all input layers are listed in Table 1 and range from 30 m (land cover) to coarser gridded products for climate variables. To combine these into a single multi-criteria surface, all layers were resampled onto a common analytical grid at 0.02° (~2.2 km at the equator), generated programmatically over Indonesia's national boundary, with each input layer reprojected onto this grid using bilinear resampling. This resolution was selected as a practical compromise: it is fine enough to preserve meaningful spatial heterogeneity in the higher-resolution layers (e.g., land cover, elevation) at sub-provincial scale, while remaining coarse enough to be computationally tractable for a continuous national-extent analysis.

This choice has two consequences worth noting. For the fine-resolution inputs like land cover and elevation, moving from 30 m to 2.2 km means each output cell is now an average of many original pixels. We gain computational tractability at national scale, but we lose the fine-grained detail that existed in the original data. For the coarser climate inputs, the opposite problem applies. Bilinear interpolation fills in smooth values between the original, more widely-spaced data points. These interpolated values give the appearance of finer spatial detail, but they don't represent new measurements. They are mathematical estimates derived from the coarser original grid, and should be interpreted accordingly.
Line 192: We thank the reviewer for recognizing this analysis as a valuable contribution. The correlation analysis was conducted at the pixel level, using the standardized Ecological Benefit Index and the Disease Risk Index as the two variables. Prior to analysis, the ecological benefit raster was reprojected and resampled to match the spatial resolution and extent of the disease risk layer using bilinear interpolation, ensuring spatial alignment between the two datasets. Valid pixel pairs from both layers were then extracted and restricted to wetland areas only (n = 21,949,553 pixels), and the Pearson correlation coefficient was computed using Python (numpy library), with statistical significance assessed at p < 0.001.
Lines 265 – 270: We thank the reviewer for pointing this out. We completely agree that socio-economic details like preference for private clinics and specific occupational exposures are not inputs to our spatial MCDA. The factors referenced in Lines 265–270 are intended to provide a broader contextual explanation for the risk patterns observed in Papua, drawing on existing literature about the region, rather than claiming these are direct outputs of the MCDA. Our intention in this paragraph was to use these real-world observations to contextualize and validate why the high-risk outputs of our environmental model align with the persistent realities on the ground in Papua.
Lines 285-288: We respectfully disagree with the characterisation that the other results are merely artifacts of author decisions. While it is true that weighted overlay analysis reflects the inputs and weights chosen, this is inherent to all MCDA approaches and does not diminish the analytical value. The explicit, transparent integration of criteria spanning the disease risk and ecological benefit into a spatially explicit conservation planning tool represents a methodological contribution in itself. Furthermore, we note that the absence of formal stakeholder elicitation is a recognised limitation of desktop GIS-MCDA studies, and is explicitly acknowledged in our limitations section.
Line 314: To clarify, this conclusion is directly based on the visual consistency of the scenarios in Figure 3. The scenarios look spatially similar not because of a weighting bias or rigid framework, but because of the empirical data reality. As shown in the individual maps, regions like Papua exhibit extremely high Ecological Benefit scores and extremely high Disease Risk scores simultaneously. Mathematically, if a pixel has a high Ecological score and a high Risk score, any combination of scenario weights will result in a consistently high final priority score. Any valid model combining these dimensions should consistently flag Papua regardless of weight shifts. Our framework successfully captured this real-world convergence, which is exactly what we meant by “capturing fundamental conservation-health relationships.” Furthermore, regarding 'flexibility': while the overarching spatial pattern remains stable, the scenarios primarily change the intensity of the priority scores, especially in moderate-risk areas like Sumatra and Kalimantan. This allows different institutions to adjust their specific intervention thresholds based on their policy leanings.
Line 320: We respectfully disagree that the analysis amounts to a trivial combination of information. We would point the reviewer to the Introduction (lines ~70–78), where we identify the specific gap this study addresses: current Indonesian conservation policy lacks systematic tools to weigh ecological benefits against public health risks, and existing approaches treat spatial variability in disease exposure and conservation value separately rather than jointly. Spatial MCDA is indeed an established method that used elsewhere to formalize vector-borne disease risk assessment (e.g., Hongoh et al., 2011). And its established status is precisely what makes it a credible vehicle here, not a weakness. What is novel is not the MCDA technique itself but its application: the systematic integration of ecological conservation value and vector-borne disease risk into a single spatial decision framework, which to our knowledge has not previously been done for wetlands at a national scale.

We would also push back on the characterization of this as merely "combining information." The MCDA framework requires translating qualitative, literature-derived expert judgment into a quantitative, reproducible, spatially explicit structure, combined transparently with objective remote-sensing and socio-economic data. That formalization is itself the methodological contribution: it converts an otherwise implicit, case-by-case planning judgment into an explicit, auditable, and transferable framework. Critically, the key finding that ecological value and disease risk are largely spatially independent (r = 0.099, p < 0.001) could not have been anticipated without the analysis, it is not an assumption we could derive beforehand. This result has direct practical implications: it indicates that, conservation planning in most regions does not need to be systematically constrained by disease-risk trade-offs.
When we refer to "framework", we mean the structure (hazard, exposure, vulnerability, and ecological benefits combined through spatial MCDA), not a fixed set of weights. We are not claiming the Indonesia-specific weights have been validated elsewhere, and we agree that would require empirical testing we have not done.
Most of the underlying data are global, open-access products (Malaria Atlas Project, SRTM, Copernicus/CRU, WorldPop…), so the same structure and data pipeline could be applied to other regions. However, the weights and value scales would need to be re-elicited and validated locally rather than reused directly.
Line 423: We agree that the individual components of our methodology, including weighted overlay analysis, spatial MCDA, and open-access datasets, and not in themselves novel. The contribution of this study is not the invention of a new method, but rather the novel application of spatial MCDA to simultaneously address wetland conservation and vector-borne disease risk at a national scale in a data-scarce tropical context. To our knowledge, no prior study has systematically integrated the hazard-exposure-vulnerability disease risk framework with ecological benefit assessment for wetland conservation planning at this scale in Indonesia or comparable tropical settings. We will revise line 423 to remove "breakthrough" and frame this more precisely as a cross-disciplinary integration rather than a methodological innovation.

Response to the textual comments:

Line 64: We agree and will revise "framework" to "solutions" at this point to align with standard NbS terminology.
Line 72: We thank the reviewer and will add this citation along with 2–3 representative examples from the spatial MCDA literature.
Figure 1: We will add a legend clarifying the color coding. Phases 2 and 3 are separated to distinguish data preparation from the MCDA aggregation steps.
Line 123: We will revise our terminology to "value scaling" for consistency. We note that "normalization" is also widely used in the broader GIS-MCDA literature (including several of the comparator studies in the newly added Table), but we agree that “value functions” terminology is more precise.
Line 133: Agreed, and we will revise this passage to accurately reflect the compensation spectrum rather than treating it as a binary.
Equations: We thank the reviewer for this comment. We will revise our notation to clearly separate criterion-level and domain-level weights as follows: the Disease Risk Index will be defined as R = Σ(wᵢ,ᵣ × sᵢ,ᵣ) for i = 1...8 risk criteria, and the Ecological Benefit Index as E = Σ(wᵢ,ₑ × sᵢ,ₑ) for i = 1...3 ecological criteria, where wᵢ and sᵢ denote the weight and standardized score of individual criterion i within each domain. The final domain-level (scenario) weights, denoted Wᵣ and Wₑ, are then used only in the final combination: P = Wₑ × E + Wᵣ × R, where Wₑ + Wᵣ = 1. We believe this revised notation more clearly conveys the hierarchical structure of the aggregation and will improve the reproducibility and clarity of the methodology section.

Citation: https://doi.org/10.5194/egusphere-2026-94-AC2

Yiting Zhu, Marleen de Ruiter, Sophie Buijs, and Nicole van Maanen

Viewed

Total article views: 1,560 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
1,109	328	123	1,560	136	125

HTML: 1,109
PDF: 328
XML: 123
Total: 1,560
BibTeX: 136
EndNote: 125

Views and downloads (calculated since 15 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	432	120	70	622
Feb 2026	196	86	10	292
Mar 2026	357	83	31	471
Apr 2026	66	26	5	97
May 2026	51	8	3	62
Jun 2026	7	5	4	16
Jul 2026	0

Cumulative views and downloads (calculated since 15 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	432	120	70	622
Feb 2026	196	86	10	292
Mar 2026	357	83	31	471
Apr 2026	66	26	5	97
May 2026	51	8	3	62
Jun 2026	7	5	4	16
Jul 2026	0

Viewed (geographical distribution)

Total article views: 1,565 (including HTML, PDF, and XML) Thereof 1,565 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 02 Jul 2026

Short summary

Restoring wetlands mitigates climate change, yet fears persist that they spread mosquito-borne diseases. We mapped environmental data across Indonesia to test this. Surprisingly, we found no significant link between high conservation value and disease risk. This proves that protecting nature does not necessarily endanger public health. Our results identify safe areas for conservation, demonstrating that we can safeguard both ecosystems and human communities simultaneously.


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