Use of Unmanned Aerial Vehicles for Building a House Risk Index of Mosquito-Borne Viral Diseases

Round 1

Reviewer 1 Report

This study provides a proof of concept of the use of unmanned aerial vehicle technology as a possibility to collect spatial information in real-time of the landscape through multispectral images for the generation of a predictive model that allows establishing a risk index relating socio-demographic variables with the presence of the vector in its different larval, pupal and adult stages. The authors developed a methodology based on partial least squares (PLS) which takes into account the different type of variables are involved, and the geographic distribution of the houses as well. Results show spatial pattern of low-risk housing at downtown which increases as we approach the outskirts of the town. The predictive model of dengue transmission risk developed through orthomosaics can help decision-makers to plan control and public health activities. 

However, the authors should explain the following issues:

1. The authors do not introduce the challlenging and difficulties of this filed. Besides, you do not introduce your approach framwork and contributions in introduction part.

2. What makes your approach unique comparing with other methods?

3. You mention that UAV can do something that the satellite cannot do. But in this paper, you do not explain why your data that satellite can not access OR you compare the results that the data access by satellite.

4. More experiments comparison should add in the evaluation part.

5. No conclusion part in this paper.

Author Response

We appreciate the reviewer's comments that enriched the manuscript.

 

Reviewer 1.

 

1. The authors do not introduce the challlenging and difficulties of this filed. Besides, you do not introduce your approach framwork and contributions in introduction part.

R= We include a paragraph on line 67-76.

Nonetheless, for the best performance of the UAV, it is necessary to consider the meteorological conditions and know the flight area [41, Thibbotuwawa et al. 2020] and the terrain's characteristics, such as high trees and telecommunications antennas, to avoid electromagnetic interference [41]. As a user of UAVs, the national regulations must be observed [46]. Also, it should be considered that one of the factors for UAV flight coverage depends on the battery time [3, Thibbotuwawa et al. 2020]. Indeed, before the flights, the field teams should visit and talk with the health authorities, principal neighborhood actors, and community leaders, which are necesary for developing these new tools [Hardy et al. 2022]. The costs of this technology must be taken into account and considered as open-source software for affordable technology. A multidisciplinary team is also required to address the entire strategy.

 

 

2. What makes your approach unique comparing with other methods?

Para Víctor.

 

R= We thank the reviewer for this comment, which allows us to highlight one of the main aspects of our proposal.

When we searched for similar works in the literature (based on Machine Learning approaches), we found a drawback in the way independent variables are used, particularly when they are mixed-type data. The proposals we found, addressed this problem mainly as a classification one, which implies a categorization of the dependent variable (e.g., high and low mosquito abundance). Although it is a good option in terms of prediction, our aim is to get an insight into the local characteristics of the area of study, which makes it more prone to mosquito-borne diseases.

To clarify this idea, we added a paragraph in Section 2 (Related Work), where we stated that “Those approaches are interesting in terms of prediction, and the use of ML models provides flexibility because most of them are non-parametric and do not assume apriori any probability distribution of data. However, for us it is very important to gain some insight into the local characteristics of the area of study, which makes it more prone to mosquito-borne diseases. A study of the importance of the variables can be done based on the parameters of the models (\cite{Cianci2015}, \cite{rahman:2021}), when it is possible, correlation analysis, or by specific hypothesis tested on the models (\cite{Chen2019}), based for example, on manually adding or excluding individual or sets of variables (\cite{rahman:2021}). All those approaches must be carried out carefully when our variables are mixed-type data, particularly for ML models based on Euclidean distance between observations. For this reason, in our approach we pay special attention to obtaining an adequate distance metric for mixed-type data, that allows us to obtain an index representing the local characteristics of houses which are related in some way, to dengue disease.”

In the same Section 2, we highlight the problem we found with the proposals we found in the literature that are similar to us, in the sense that they used ML methods as well.

About \cite{Chen2019}, we said that: “there are no further details about the distance metric used in the binary input variables case, because, except for random forest, all classification methods [k-nearest neighbor, support vector machines, neural networks] use Euclidean distance as the default option, even when a non-linear kernel or activation function is used, i.e., they are originally formulated for continuous input variables.”. Respecto to the proposal of \cite{rahman:2021}, we state that: “it is not clear how the mixed input data (categorical and continuous) are managed, because all classification models they used except random forest, cannot handle mixed type data naturally. Maybe it is the reason why random forest showed the best results.”

Finally, in Section 5, Discussion, we emphasize that “Certainly, there could be many advantages of incorporating more information from many sources, because they could enrich our knowledge of the phenomenon we are interested in, and arguably, could improve the model we used for solving some specific task; however, it is very important to pay attention in the way this information is used by our models, especially when different structures are present, giving us mixed-type variables. We think the latter is crucial, particularly for models where a distance function among objects (houses, in our case) is involved, as is the case of ML-related models, because not all distance functions are capable of representing differences between objects with mixed-type data. Consider for example, the Euclidean distance, which is used by default in most of the supervised and non-supervised ML algorithms, and is aimed at continuous (real-valued) variables, then, if we use a model based on Euclidean distance for categorical or mixed-type data, the results and conclusions we obtain with this model, may not be correct. In our proposed index, we take into account the different types of variables by using an efficient encoding procedure which gives us a shared metric space with a corresponding norm capable of representing distances between any two objects taking into account the two types of input variables we used, and we think this is one of our main contributions, given that similar works reported in the literature, it is unclear how the mixed input data (categorical and continuous) are managed, or the distance function they used.”

 

 

3. You mention that UAV can do something that the satellite cannot do. But in this paper, you do not explain why your data that satellite can not access OR you compare the results that the data access by satellite.Discutido:

 

R= We include two paragraphs on line 487-192.

Nevertheless, the satellite images are taken at different seasons, daily times, and camera angles, which does not allow the precise association of the environmental variables related to the Ae. aegypti house biological cycle. In this study, the UAV photos were taken at 100 m, around noon, and simultaneously with house surveys [12, 48 y 49] making it possible to associate the vector's biological activity with landscape elements since it allows us to determine them precisely and according to the field data record.

 

4. More experiments comparison should add in the .

 

R= Once again we thank the reviewer, we think that her/his request allowed us to explore more characteristics of our model, and to gain knowledge about the effect of the parameters.

Initially, we wanted to explore the performance of our methodology as a “feature-generation” model (with the Factor Analysis for Mised Data approach we used), to address a classification problem, in the same way other researchers did, however, we would need to restrict to a binary classification problem only, in order to avoid imbalance categories, and this is not our purpose. So, we decided to explore different parameterizations of our model, that we consider could give us a better knowledge of the effect of those parameters for the risk index we propose.

We added a paragraph in the beginning of Section 4, Results, where we explain on that:

“We performed an extensive set of experiments in order to explore interesting patterns and to select a good model to obtain our proposed index. Those experiments where carried out based on a grid of parameters, which we consider important, and are described on Table 2.

As any regression-based model, in PLS we can obtain some metrics accounting for its predictive performance, however, it is not our objective (it is very easy to show that a performance metric such as the coefficient of determination, can be artificially "inflated", resulting in an overfitted model). In our case, we want to take advantage of an unique characteristic of PLS: to construct the best set of latent variables (components) based on the set of factors we obtained with FAMD,  which are correlated to our response variable of interest, as we explained in Section \ref{sec:index_pls}. In this sense, we want to exploit the explanatory habilities of PLS which will allows us to obtain a set of components that can represent all covariates we used (from different sources and types) in an useful way to be considered as a risk index.

Then, our main evaluation criteria was the explained inertia (covariance) obtained with the FAMD procedure according to the $PC_{FAMD}$ variable, and the residuals on log counts of the response variables. Because we have many possible combinations of the parameters, we will not show here those experiments, but many of them can be found as supplementary material to this paper in https://github.com/victorm0202/mosquito\_borne\_viral\_disease\_paper.”

Then, we mention some of the relevant characteristics we found:

“We observe that the patterns of risk we found, are consistent in most of the parameterizations we used, except on those where the model overfits, and we observed also, that this tends to occur when PC_{PLS} parameter is high (5 or 6, mainly). In this case, the risk index does not show a particular spatial trend, even when we vary the other parameters. “

Based on that, we choose an adequate set of parameters (we added the Table 2 to explain the main parameters of the model), which we used to show all the results.

We found this exercise very interesting. Thank you very much again for the comment.

 

 

5. No conclusion part in this paper.

R= We include two paragraphs in line 614-619

This study shows that the use of UAVs can be incorporated into vector surveillance and control strategies, providing spatial and temporal data in real-time of the landscape components that allow dengue vectors to continue transmitting the disease. On the other hand, the predictive mathematical model proposed in this study to estimate the risk of dengue transmission demonstrates high reliability in identifying high-risk areas within the study locality and managing control activities specifically.

Reviewer 2 Report

This manuscript is interesting and the topic is suitable to be published in the special issue. The description in the content is very detailed and readable. Please authors consider the following terms for revision.

Please authors brief describe the counting approach and how to identify the eggs, larvae and pupae of Ae. aegypti, especially if the different kinds of mosquitos, not Ae. aegypti, were available in the containers, how to count the number of mosquitos? Is it necessary that the assistant, who implemented counting of eggs, larvae and pupae of Ae. aegypti, should be educated or trained? Additionally, in Ref. 40, the letters of one author’s name are all capital, and it should be corrected.

Please authors further state that the technology described in this study how to be applied to the real community in the future.

Author Response

We appreciate the reviewer's comments that enriched the manuscript.

 

Reviewer 2.

 

Please authors brief describe the counting approach and how to identify the eggs, larvae and pupae of Ae. aegypti, especially if the different kinds of mosquitos, not Ae. aegypti, were available in the containers, how to count the number of mosquitos? Is it necessary that the assistant, who implemented counting of eggs, larvae and pupae of Ae. aegypti, should be educated or trained?

 

R= We made modifications in the section Entomological survey for adults of Ae. aegypti, specifically in line 196-199.

in the field and recorded per house. Then, the adult mosquitoes were transported in small cages covered with mash and labeled, one per house, into a cooler to the laboratory. The adults were then identified morphologically according to (Darsie and Ward 2016).

 

In addition, the line 207-213 was added.

Ladles, mesh nets, mosquito larvae dippers, and plastic pipettes were used to check each container. The small and medium containers were emptied for larvae and pupae counted, and for the large containers, the personnel used small containers where all the immature stages of Ae. aegypti were counted by depth until no larvae or pupal activity was observed in the container [40]. The species were identified in the field according to (Darsie and Ward 2016). Previous training was necessary for field personnel to strengthen their knowledge about the vector and fill out the entomological surveys.

 

Additionally, in Ref. 40, the letters of one author’s name are all capital, and it should be corrected.

R= Changed

 

Please authors further state that the technology described in this study how to be applied to the real community in the future.

Finally, the implementation of this methodology has as its final objective the inclusion for improvements in the community. We know that due to the pandemic, the interaction measures between the staff of the health secretary and the community must be respecting the distance between them. Therefore, by using this methodology, apart from having new information that will improve decision-making, we have a perspective of inclusion in the community.

 

Round 2

Reviewer 1 Report

The authors have revised the paper according to the suggestions.