Effect of El Niño Southern Oscillation cycle on the potential distribution of cutaneous leishmaniasis vector species in Colombia

Local anomalies in rainfall and temperature induced by El Niño and La Niña episodes could change the structure of the vector community. We aimed to estimate the effect of the El Niño–La Niña cycle in the potential distribution of cutaneous leishmaniasis (CL) vector species in Colombia and to compare the richness of the vectors with the occurrence of CL in the state of Norte de Santander. The potential distributions of four species were modeled using a MaxEnt algorithm for the following episodes: La Niña 2010–2011, Neutral 2012–2015 and El Niño 2015–2016. The relationship between the potential richness of the vectors and the occurrence of CL in Norte de Santander was evaluated with a log-binomial regression model. During the El Niño 2015–2016 episode, Lutzomyia ovallesi and Lutzomyia panamensis increased their distribution into environmentally suitable areas, and three vector species (Lutzomyia gomezi, Lutzomyia ovallesi and Lutzomyia panamensis) showed increases in the range of their altitudinal distribution. During the La Niña 2010–2011 episode, a reduction was observed in the area suitable for occupation by Lutzomyia gomezi and Lutzomyia spinicrassa. During the El Niño 2015–2016 episode, the occurrence of at least one CL case was related to a higher percentage of rural localities showing a richness of vectors = 4. The anomalies in rainfall and temperature induced by the episodes produced changes in the potential distribution of CL vectors in Colombia. In Norte de Santander, during Neutral 2012–2015 and El Niño 2015–2016 episodes, a higher probability of at least one CL case was related to a higher percentage of areas with a greater richness of vectors. The results help clarify the effect of the El Niño–La Niña cycle in the dynamics of CL in Colombia and emphasize the need to monitor climate variability to improve the prediction of new cases.

Minor changes in text and in layout of figures should be considered (details are in comments in the attached pdf). R: We made the suggested changes in the attached pdf, see bellow.
Reviewer #2: Please see Summary and General Comments Reviewer #3: The analyses are adequate and the results are quite clear. The figures are adequate and sufficient. It is recommended that the explanation in the text of the Figures in general should be done in the order in which they appear (A, B, C) and/or be consistent with the models are presented in Materials and Methods: neutral episode, "La Niña" episode and finally "El Niño" episode. R: We reorganized the next paragraphs: Lines 403-411 "Of the 1,705 localities identified in Norte de Santander, 40 (2.3%) are urban areas, and 77 (4.5%) are small villages. Among the 1,588 rural localities, during the Neutral episode of 2012-2015 (Fig 6A), 10.8% (172 localities) presented CL cases (372 cases in total, with a median 0 per locality, and ranging from 1 to 24 cases per locality when cases were present); during the La Niña episode of 2010-2011, only 5% (79 localities) had CL cases (131 cases in total, with a median of 0 per locality, and ranging from 1 to 17 cases per locality when cases were present) ( Fig 6B); and during El Niño episode of 2015-2016, 12% (191 localities) had CL cases (511 cases in total, with a median of 0 per locality, and ranging from 1 to 19 cases per locality when cases were present) ( Fig 6C)." Lines 431-440 "These statistical findings are related to the observations presented in Figure 7. The intersection between rural localities with potential richness of vectors ≥3 and rural localities with at least one CL case during the Neutral 2012-2015 episode was lower and corresponded to the center zone of the state (Fig 7A). The lowest intersection between rural localities with potential richness of vectors ≥ 3 and rural localities with at least one CL case corresponded to the La Niña 2010-2011 episode ( Fig 7B). During the El Niño 2015-2016 episode, nearly the entire state of Norte de Santander (except the northeast region) presented a potential richness of vectors ≥3, and the intersection with rural localities with at least one CL case represented an important extension in the southern and eastern regions of the state ( Fig 7C)."

Conclusions
Reviewer #1: The findings is worthy to be published. The discussion section, as well as the Introduction section, are well written and substantiated. The limitations and necessary advances about the study were pointed out.

Reviewer #2: Please see Summary and General Comments
Reviewer #3: The limitations of the work are well delimited and they made relevant approaches to solve them losing spatial resolution but that served to mark a trend. The main problem in the areas affected by CL is the recording of the case because of the "time span" until the symptoms appear, which makes it difficult to get the date and probable place of infection of this pathology.
R: We agree with the reviewer. We added in lines 524-526 "Likewise, it was difficult to establish the date when the transmission occurred, and the symptom onset date was not available, so the date when the case was recorded by surveillance system was used."

Editorial and Data Presentation Modifications
Reviewer #1: Minor changes are indicated throughout the manuscript file (see attached pdf file). R: We made the suggested changes in the attached pdf, see bellow.
Reviewer #2: Please see Summary and General Comments Reviewer #3: "Minor Revision"

Summary and General Comments
Reviewer #1: There is no doubt about how weather conditions affect vector-borne diseases, but it is necessary to study the true impact of the recent climate changes in burden of vector-borne diseases and the authors have collaborated on that, analysing the climate effect in distribution of Lutzomyia vectors on cutaneuous leishmaniasis cases. They report the analyses based on recent El Nino -La Nina cycle, which is a short period, but is worthy to be published, because I considere that small data collection is value to further being part of larger and comprehensive studies. R: We accepted the suggestion of the reviewer, and we changed the sentence to clarify the idea we want to transmit to the readers, about the novelty of our research (lines 150-153): "In our knowledge, not exist a previous study that simultaneously appraise; a) the effect of the ENSO cycle on the potential distribution of the vectors of CL, and b) evaluate the effect of the changes in the potential distribution of vectors on the occurrence of cases of CL". 3)Methods: please provide some uncertainty measure for pROC (e.g., minimum and maximum based on bootstrapping). The problem of partial ROC analysis to test statistical significance of ecological niche model predictions is that it does not provide a measure of how much good is the prediction, you just know that 1.0 corresponds to a random classifier, so you can test significance but this information which is too vague to assess the quality of the prediction. Even knowing the limitations of the regular AUC analysis in such a setting, I would like to see the % of correct classifications for presence points and for a random sample of absence points, and also the number of wrong presence predictions.
R: This was a useful suggestion. Now we have included in the results the edited table including the maximum and minimum values, correct classification rate and omission error (fraction). See Table in line 347. 4) Methods: The authors used a random sample of 30% of the distribution data to evaluate the model. Please make sure to give a detailed description of this process. Is the 30% used to test and validation was done in the other 70%? Did you use some cross-validation? Is the error greater or lower for points located in the state of Norte de Santander as compared to points outside?.
R: As recommended, technical details on the methodology were modified and we included in line 291-292 "…and occurrence data for each species was split into training (70%) and evaluation (30%).". We didn't use cross-validation. On the other hand, is important to clarify that the calibration of the model was carried out in the M and was transferred and evaluated in the entire study area, the political division the state of Norte de Santander was just used to evaluate the relationship between the potential richness of the vectors and CL occurrence. 5) Discussion: should highlight problems in prediction based on incomplete data (only those that could be assessed --publication bias; mainly presence data -misclassification bias; data not well distributed in the area, etc.). R: We accepted the suggestion of the reviewer, and we modified the lines 536-538 to include the bias mentioned by the reviewer. "For last, we recognize gaps in the prediction of potential distribution of CL vectors, based on incomplete data; publication, taxonomic, and misdetermination bias; and not well distributed data in the country." Reviewer #3: In general the study is well supported by the data it uses and provides in the supplements. Some minor revisions and further details are highlighted and requested in the manuscript.
Highlighted and requested in the manuscript PDF.

a. REVIEWER (1)
Line 204-208: It is explained why the reclassified coverage layer was included, but nothing is said about its influence on results and discussion. R: We understand the reviewer point, but the results section requires a brief functional analysis and we do not mention in a specific way the contribution of any variable. However, in our discussion we have explained how the land use and land cover change could be influencing our results, lines 482-484 "However, these changes were probably mediated by the land cover and the suitability of the habitat for the establishment of viable populations of vectors, particularly in forests [6,44] and perennial crops (e.g., coffee and cocoa)".