Studying Land Cover Changes in a Malaria-Endemic Cambodian District: Considerations and Constraints

Round 1

Reviewer 1 Report

This manuscript can be published as a case study of using various methods for analysis of rapidly growing deforestation in malaria area. The discussion is presented completely, although the advantages of the method proposed by the authors are not clearly described. A small note is that the authors should clarify the possible increase in malaria cases since 2016.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Please see attached review.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The revised manuscript has addressed my previous concerns.

Grammatical points:

137 makes place = takes place

182 emit hypotheses. Not sure emit is the right word. Formulate?

245 population = the population

248, 393 cultures = cultivated areas?

336 oncoming investigations in the study area. Upcoming? What are these?

468 induce = induces

476 “Hypothesis which would correlate with the recent investment into rubber and cashew trees observed in the area.” Not a proper sentence

486 independents = independent

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This is a valuable contribution to the study of land cover evolution and diseases transmission. The authors investigate the analysis of land cover changes and the progression of deforestation within one of malaria endemic area in Cambodia. The manuscript combined well-explained use of satellite processing with solid knowledge of the subject matter. I found the article quite interesting and it needs minor revision in order to be published.

More attention should be given to the spatiotemporal patterns of malaria at province-level.

L.125-128. It would be desirable if the authors provide the visual plot of malaria incidence in Cambodia and Mondulkiri province from 2008 to 2018. The series of province-level maps of malaria incidence also would be useful for assessment of malaria problem in Mondulkiri province versus other provinces. It could help a better understanding of malaria dynamics on national and province levels.

L.170-171. A brief description of social and environmental features of Kaev Seima district could provide more information about possible conditions for malaria transmission.

L.258-260. Please add the information about changes in built-up areas.

L. 321—362. Please use more spatial information about malaria incidence at province-level for discussion. Are there any other provinces with similar trends in malaria and deforestation? What is the incidence rate in Mondulkiri province (high or low)?

Reviewer 2 Report

Review of Pepey et al. 2020, Remote Sensing: “Land cover changes in Cambodia: insight on deforestation and malaria dynamics within Mondulkiri Province.”

The authors present a land use classification, focusing on deforestation and fragmentation, for a region in Cambodia that has experienced a recent increase in malaria transmission. The land use classification is certainly a potentially important contribution to future studies of malaria ecology and transmission dynamics in the region, particularly given the apparent lack of such standardized land use/land cover data for Cambodia. However, given the lack of malaria case reporting data or vector surveillance data, and lack of any analysis of land use effects on these outcomes, the study amounts to a simple land use classification with only speculation as to its relevance to malaria.

In the absence of any real analysis of land use change and malaria transmission in the study region (in contrast with the studies discussed in the introduction that the authors appear to dismiss for their shortcomings) it is hard to actually see this as a ‘malaria study’ – rather it is a land use classification in a small region of Cambodia that might be relevant for future malaria studies.

Further, despite the lack of malaria/vector data, only minimal ecological context is presented, and much of it focuses on the malaria system in Latin America. For example, what is the primary malaria vector in the study region in Cambodia? What is known about its ecology in SEA and how is that relevant to the land use classification you conduct? What about standing water? Could rice paddies be important breeding habitat for the mosquito vectors, or do they breed in forest pools? Is there earlier literature that you can discuss that explores land use effects on malaria transmission or entomological risk in the region? The ecological context is lacking.

The methodology employed, while reasonably robust, is not novel (common techniques for image classification) so it is not a novel methodological contribution (in the absence of contributing to the malaria-deforestation literature). The methodology (manual image classification) also does not seem scalable for standardized application across large regions and through time (which would be necessary for its utility in identifying effects of land use change or deforestation on malaria transmission).

I suggest the authors take one of two approaches: 1) illustrate the scalability of their method (increase spatial extent and repeat for multiple time points) to illustrate how they can detect changes in landscape structure/composition relevant to mosquito ecology and how this may be an improvement over other methods or land use classifications, or 2) conduct some actual analyses of malaria case reporting using the land use classification they have produced – such studies have been conducted in many tropical regions/countries (Indonesia, sub-Saharan Africa, Latin America) and to be a contribution to the malaria-deforestation literature, some analysis of malaria transmission in this study would be necessary.

Minor comments:

• Table 1 appears to be cut off.
• Were the fragmentation metrics (ED, PD) calculated for the forest class? Or others? Be explicit about this.
• Based on figure 9, it appears the change in fragmentation metrics in 2018 could be largely attributable to the change in methodology rather than to actual changes in the landscape, which is problematic for interpretation (particularly in regards to effects on aggregate malaria case reporting).
• The authors emphasize the importance of separating different forest types (e.g. natural forest and plantation forests), which I agree is important – I would suggest looking into the Mapbiomas project in Brazil for an example of a very large-scale (Brazil + Amazon basin, annually for multiple decades) project that has classified imagery into categories that include different forest types (e.g. natural forest, savanna, mangrove, forest plantation).
• I would suggest reading studies investigating and explaining the concept of frontier malaria (e.g. Castro et al. 2006, PNAS “Malaria risk on the Amazon frontier”; Baeza et al. 2017, Nature Ecology and Evolution “The rise and fall of malaria under land-use change in frontier regions”) as they seem to capture the malaria dynamics the authors hypothesize for Cambodia/SEA.
• The writing is reasonably clear, but sentence structure and grammar could be dramatically improved for ease of interpretation; I would suggest having a native English speaker proof read the manuscript before resubmission.

Reviewer 3 Report

The authors classify maps of a district of Cambodia in 1988, 1998, 2008, and 2018 into four land-cover classes: fields, forest, plantations, and built up areas. They further condense forest and plantation into wooded areas and show that fields have increased at the expense of wooded areas from 1988-2018 (Figure 6). Using these land-cover change data, several derived statistics related to forest fragmentation, and data on trends in malaria rates in the region (not shown in the paper), they discuss the relationship between deforestation and malaria, concluding that “it appears that malaria does not follow linearly deforestation expanse. It is important to consider that malaria is not only driven by deforestation and that temporal scale matters.”

The relationship between deforestation and malaria is an important, open topic. Many—though not all—previous studies have found a positive relationship between the two. However, the analysis presented in this paper is incomplete. The authors present data, analysis, and results only for land cover change, not for malaria. Thus any conclusions the authors draw are not based on any formal quantitative analysis shown in the paper.

I’m skeptical how much can be learned about the complex relationship between deforestation and malaria from just three decade-long time periods and aggregate-level malaria data, given the complexity of the relationship and the other confounding factors at work. But perhaps the malaria data could be spatially disaggregated sufficiently to enable a multivariate regression analysis, or another statistical analysis. Other studies of deforestation and malaria take complexity and non-linearity as a starting point and attempt to resolve the role of deforestation in malaria through multivariate regression. Perhaps the authors could build upon the present analysis to do the same.

Specific comments

74 I would question calling Austin et al a meta-analysis, since it’s not a study of studies. Rather it’s a single study that uses data from many countries.

73-95 Note that not all studies of deforestation and malaria found a positive association, e.g. Bauhoff and Busch (2020).

104 “First among other countries, Cambodia raises concern as the number of estimated malaria cases are increasing since 2017 [30].” A keyword skim of the WHO 2019 malaria report for “Cambodia” does not turn up any statistics suggesting that Cambodia is first or most severe in any malaria metric. Indeed I find “In 2018, Cambodia reported no malaria-related deaths for the first time in the country’s history.” The worthiness of studying deforestation and malaria in Cambodia can be justified without superlatives.

127 The statistics are for 2017 and 2018, but the second reference is from 2010. Misplaced reference?

141-153 This section makes the case for distinguishing more land-cover classes than forest/non-forest, but ultimately this study also aggregates forests and plantations into “wooded areas” for the purposes of analysis (line 214)

169 why was this study site selected?

175 Please describe the malaria data in more detail. Show the data and trends. And ideally conduct an analysis that uses these malaria data and the land-cover change data.

274 the Results section concludes without any comparative analysis of land-cover change and malaria

301 “However, it is worth noting that, along with edge density, decreases patch density” the meaning of this sentence is unclear; perhaps it is missing a word?

321 Over the last decades, at national level, it has been assessed and quantified that deforestation has been increasing [15,52], while malaria incidence has been decreasing up to 2016.” Whatever we are to infer from this should be shown formally and statistically in the Methods and Results, rather than introduced in the Discussion

322 “As SEA is characterized by forest malaria, we could expect an overall decrease of malaria incidence as the forest cover decreased.” The a priori hypotheses about the relationship between malaria and forest cover should be stated before the Methods. It’s not clear that this is the correct a priori hypothesis, since the literature, including as cited in the paper, suggests malaria transmission is a function of deforestation, or edge proximity, rather than total forest presence.

321- This paragraph casually discusses various data points and trends, but in the absence of any formal analysis from the paper.

340 “As the deforestation becomes more intense on an extended period, potentially eliminating most of the forest – and considering current pace, it most probably will” This speculation should be justified with a reference or analysis.

349 “It is important to consider that malaria is not only driven by deforestation and that temporal scale matters” Yes, for this reason many other studies take into account other variables through multivariate regression analyses. Such methods aren’t perfect, but they can go farther toward modeling a complex phenomenon such as malaria and deforestation than univariate analysis.