UBC Faculty Research and Publications

Predicting the geographical distributions of the macaque hosts and mosquito vectors of Plasmodium knowlesi… Moyes, Catherine L; Shearer, Freya M; Huang, Zhi; Wiebe, Antoinette; Gibson, Harry S; Nijman, Vincent; Mohd-Azlan, Jayasilan; Brodie, Jedediah F; Malaivijitnond, Suchinda; Linkie, Matthew; Samejima, Hiromitsu; O’Brien, Timothy G; Trainor, Colin R; Hamada, Yuzuru; Giordano, Anthony J; Kinnaird, Margaret F; Elyazar, Iqbal R F; Sinka, Marianne E; Vythilingam, Indra; Bangs, Michael J; Pigott, David M; Weiss, Daniel J; Golding, Nick; Hay, Simon I Apr 28, 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-13071_2016_Article_1527.pdf [ 2.23MB ]
Metadata
JSON: 52383-1.0366939.json
JSON-LD: 52383-1.0366939-ld.json
RDF/XML (Pretty): 52383-1.0366939-rdf.xml
RDF/JSON: 52383-1.0366939-rdf.json
Turtle: 52383-1.0366939-turtle.txt
N-Triples: 52383-1.0366939-rdf-ntriples.txt
Original Record: 52383-1.0366939-source.json
Full Text
52383-1.0366939-fulltext.txt
Citation
52383-1.0366939.ris

Full Text

RESEARCH Open AccessPredicting the geographical distributions ofthe macaque hosts and mosquito vectorsof Plasmodium knowlesi malaria in forestedand non-forested areasCatherine L. Moyes1*, Freya M. Shearer1, Zhi Huang2, Antoinette Wiebe1, Harry S. Gibson2, Vincent Nijman3,Jayasilan Mohd-Azlan4, Jedediah F. Brodie5,6, Suchinda Malaivijitnond7, Matthew Linkie8, Hiromitsu Samejima9,Timothy G. O’Brien10, Colin R. Trainor11,12, Yuzuru Hamada13, Anthony J. Giordano14,15, Margaret F. Kinnaird10,Iqbal R. F. Elyazar16, Marianne E. Sinka2, Indra Vythilingam17, Michael J. Bangs18,19, David M. Pigott20,Daniel J. Weiss2, Nick Golding1 and Simon I. Hay20,21AbstractBackground: Plasmodium knowlesi is a zoonotic pathogen, transmitted among macaques and to humans byanopheline mosquitoes. Information on P. knowlesi malaria is lacking in most regions so the first step to understand thegeographical distribution of disease risk is to define the distributions of the reservoir and vector species.Methods: We used macaque and mosquito species presence data, background data that captured sampling bias inthe presence data, a boosted regression tree model and environmental datasets, including annual data for land classes,to predict the distributions of each vector and host species. We then compared the predicted distribution of eachspecies with cover of each land class.Results: Fine-scale distribution maps were generated for three macaque host species (Macaca fascicularis, M. nemestrinaand M. leonina) and two mosquito vector complexes (the Dirus Complex and the Leucosphyrus Complex). TheLeucosphyrus Complex was predicted to occur in areas with disturbed, but not intact, forest cover (> 60 % tree cover)whereas the Dirus Complex was predicted to occur in areas with 10–100 % tree cover as well as vegetation mosaics andcropland. Of the macaque species, M. nemestrina was mainly predicted to occur in forested areas whereas M. fasciculariswas predicted to occur in vegetation mosaics, cropland, wetland and urban areas in addition to forested areas.Conclusions: The predicted M. fascicularis distribution encompassed a wide range of habitats where humans are found.This is of most significance in the northern part of its range where members of the Dirus Complex are the mainP. knowlesi vectors because these mosquitoes were also predicted to occur in a wider range of habitats. Our resultssupport the hypothesis that conversion of intact forest into disturbed forest (for example plantations or timberconcessions), or the creation of vegetation mosaics, will increase the probability that members of the LeucosphyrusComplex occur at these locations, as well as bringing humans into these areas. An explicit analysis of disease risk itselfusing infection data is required to explore this further. The species distributions generated here can now be includedin future analyses of P. knowlesi infection risk.Keywords: Species distribution model, Deforestation* Correspondence: catherinemoyes@gmail.com1Spatial Ecology & Epidemiology Group, The Big Data Institute, Li Ka ShingCentre for Health Information and Discovery, University of Oxford, OxfordOX3 7BN, UKFull list of author information is available at the end of the article© 2016 Moyes et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Moyes et al. Parasites & Vectors  (2016) 9:242 DOI 10.1186/s13071-016-1527-0BackgroundApproximately one million malaria cases were reportedby countries in the Indochinese Peninsula, Malay Peninsulaand insular southeast Asia in 2013, and the Plasmodiumknowlesi parasite was the most common cause of malariain Malaysia [1]. It is a zoonotic disease that can cause se-vere symptoms and fatalities in humans [2], and is trans-mitted among macaques and to humans by anophelinemosquitoes [3]. Outside Malaysia, human cases have beenreported from a small number of dispersed locations [3, 4]but the distribution of P. knowlesi in these countries islargely unknown. There are far more reports of macaqueand mosquito populations, which provides an opportunityto use these distributions to refine estimates of the geo-graphical distribution of knowlesi malaria.A World Health Organization consultation concludedthat this disease is a public health problem that is limitedto population groups that live, work in or visit forestedareas [5] and it is commonly cited as such [6–10]. Nostudy has, however, analysed the relationship between for-est cover and the distributions of the primary P. knowlesihost or vector species, across the ranges of these species,limiting our ability to understand the risk factors for dis-ease transmission.Plasmodium knowlesi parasites regularly infect Macacafascicularis and M. nemestrina macaques [11, 12] and de-tailed molecular studies have shown that recent humaninfections in Malaysia match two distinct populations ofparasites found in M. nemestrina and M. fascicularis re-spectively [8, 13]. Macaca leonina is a close relative of M.nemestrina that has only recently been classified as aseparate species [14, 15]. The distribution of M. leoninaextends further north than either M. fascicularis or M.nemestrina to areas of north Myanmar where P. knowlesicases have been found [16, 17]. For this reason, M. leoninahas been included with M. nemestrina on previous mapsof P. knowlesi risk that display overlapping ranges of thespecies involved [3] and is considered a putative hostspecies.There is good evidence that P. knowlesi is transmittedto humans by a number of mosquito species from theLeucosphyrus Group: Anopheles dirus [18] and An. cracens[19] in the Dirus Complex, and An. latens [20], An. balaba-censis [21] and An. introlatus [22] from the LeucosphyrusComplex. Earlier studies did not identify mosquitoes to thespecies level using molecular methods but they add to thebody of evidence that members of the Leucosphyrus Grouptransmit this malaria from monkeys to humans [23]. In-deed, no mosquito species outside the LeucosphyrusGroup has so far been implicated by studies conductedin the field.Previous approaches that superimposed potential hostand vector species range maps [3, 4] do not provideinsight into the variation in P. knowlesi infection riskwithin these ranges and do not provide an evidence basefor the potential link between forest cover and diseaserisk. We used species distribution models to investigatethe distributions of each of the known and putativehosts and vectors of P. knowlesi parasites, and to exploretheir relationships with forest cover, forest use, and otherrarely considered but potentially influential environmentalvariables. Our aim was to produce predicted species distri-butions, based on empirical data, that could be used infuture studies, combined with what data there is on P.knowlesi infections at different locations, to predict geo-graphical variation in disease risk in future studies.MethodsSummaryWe used a boosted regression tree (BRT) species distri-bution model, constructed in R, to examine the relation-ship between the occurrence of each macaque andmosquito species and 19 environmental covariates, andto predict the relative probability of occurrence for eachspecies at every square (pixel) in a 5 × 5 km grid. Thedata used by the model were (i) occurrence data pointsfor each species; (ii) survey location datasets that de-scribed the sampling bias in the occurrence data; and(iii) a suite of environmental variables. During the studyperiod (1990 to 2014), deforestation has led to dramaticchanges in forest cover so we constructed annual datalayers for each land cover class. Finally, the model out-puts on the islands of the archipelago were masked by arange defined for each species.Species occurrence data collationFor each species investigated, we undertook a wide litera-ture search for reports of occurrence and then appliedinclusion criteria to ensure the data quality met our mini-mum standards.For each macaque species, we conducted a Web ofScience bibliography search using the species name(including common names: long tailed macaque, crabeating macaque, kra macaque, pig tailed macaque). Wesearched the resulting articles for (i) reports of the speciesfound at specific locations and (ii) citations for othersources of occurrence data. We also wrote to the studyauthors to request unpublished datasets. Finally, we askedconservation organisations working in the region for theirunpublished data.Inclusion criteria for the macaque occurrence data were(i) reports from a specified date on or after 1 January 1990and ideally on or after 1 January 1999 to match, as closelyas possible, the year ranges for the covariate data; (ii) re-ports from a specific location representing an area notgreater than 5 × 5 km; (iii) individual species identified; and(iv) reports of free-living macaques not captive animals.Aggregated data from multiple time periods or multipleMoyes et al. Parasites & Vectors  (2016) 9:242 Page 2 of 12sites were disaggregated to single time periods and sites.Duplicate records of the same species found at the samesite within the same year were removed, with a single rec-ord for that year retained, in order to mitigate against over-sampling at specific sites. An annual time period waschosen to match the land cover data, which were generatedfor every year.Coordinates provided by data contributors were con-verted into decimal degrees. Sites without coordinatesfrom the data provider were assigned coordinates byidentifying the site in at least two online gazetteers(GeoNames, Google Earth, Google Maps, iTouchMapand/or OpenStreetMap).The mosquito data collation mirrored the process formacaque data collation above and has been reportedpreviously [24]. Low volumes of data were available formost species so we also collated data for the relevantspecies complexes (the Leucosphyrus Complex and theDirus Complex). The previous mosquito data collec-tion was extended to include reports published up toNovember 2015.In total, we collated 1,116 occurrence records between1999 and 2014 for M. fascicularis and 1,025 for M.nemestrina. The inclusion criteria that locations shouldnot be greater than 5 × 5 km was relaxed for M. leoninaand we collated 450 records from 1992 to 2014, of which33 were linked to locations representing areas > 25 km2.The borders of each area > 25 km2 were defined inArcMap.We collated 545 records for the Dirus Complex (in-cluding 107 for An. dirus and 19 for An. cracens) and49 for the Leucosphyrus Complex (including 21 for An.balabacensis, 11 for An. latens, and 9 for An. introlatus)from 1982 to 2013.Collectively the surveys used did not sample a repre-sentative set of environments at the same frequency thateach environment occurs within the study area, for ex-ample camera traps used by conservationists are rarelyset up in urban areas or impenetrable jungle. Our aimwhen collating the background datasets was to accountfor as much of the sampling bias in the presence data aspossible by selecting datasets that used the same methods(e.g. camera traps and direct observations from transectwalks) to record similar species (primates and other mam-mals). The surveys that provided presence data for themacaque and anopheline species frequently reported morethan one species and met our criteria for background dataof using the same methods to record similar species.We obtained the locations of all mammal surveys held bythe Global Biodiversity Information Facility (www.gbif.org)that were conducted on a specified date from 1990 on-wards, at a specified location and within our area of study.We also used the records from the other two macaque spe-cies to generate additional background data for the speciesbeing modelled. Each background dataset was restricted tothe range (plus a 300 km buffer to allow for uncertainty inthe ranges) of the species being modelled.To account for the sampling bias in the anopheline data-sets, we used a database of all published malaria vector sur-veys held by the Malaria Atlas Project (www.map.ox.ac.uk/explorer). Each background dataset was restricted to therange (plus a 300 km buffer to allow for uncertainty in theranges) of the species, complex or group being modelled.Table 1 provides the total number of presence andbackground data points available for each macaque andmosquito species and complex. The full distributions ofthese datasets, in space and time, are shown in Additionalfile 1.Covariate data surface constructionNineteen environmental and human population variableswere tested in the species distribution models and aresummarised in Additional file 2. The process of con-structing environmental data layers from MODIS satel-lite data has been described previously [25] and wasextended from Africa to the rest of the world, coveringthe period 2000 to 2014. In addition, to include season-ality in temperature and moisture/vegetation indices, thestandard deviation of the monthly values at each loca-tion was calculated. Of the 19 land cover classes, six thatwere not relevant to the study area were excluded. For-est data layers were constructed separately (see below)and the urban class was excluded because we used humanpopulation density data, leaving a total of seven classes(open shrubland, woody savannah, savannah, grassland,permanent wetlands, cropland, and cropland-natural vege-tation mosaic).Annual intact and disturbed forest data surfaceconstructionTotal forest cover was defined annually from 2001 to2012 by combining the five forest classes available in theInternational Geosphere and Biosphere Programme (IGBP)land cover dataset produced using MODIS satellite im-agery (MCD12Q1) [26]. We divided the aggregated forestcover data into two sub-classes defined previously by forestTable 1 Number of data points used in each modelNo. presence points No. background pointsM. fascicularis 1, 116 2, 267M. nemestrina 1, 025 608M. leonina 450 1, 041An dirus 107 447Dirus Complex 545 881Leuco. Complex 49 913Leuco. Group 615 1, 802Moyes et al. Parasites & Vectors  (2016) 9:242 Page 3 of 12landscape researchers [27]; intact forest and disturbedforest. The Intact Forest Landscape (IFL) map for the year2000 is a publicly available vector dataset encompassingareas defined as ‘an unbroken expanse of forest showingno signs of human activity and having an area of at least500 km2’ [27], which we converted to a 500 m spatial reso-lution raster dataset. The IGBP forest cover data for 2001was divided between the two sub-classes using the IFLdataset. The IGBP forest cover data for subsequent yearswas then used to reclassify any cells in the preceding yearthat had been considered forest (intact or disturbed) tonon-forest, if the corresponding IGBP cell showed a transi-tion from forested to non-forested land cover from oneyear to the next. If the IGBP data showed a transition fromnon-forest to forest, the cell was defined as disturbed forestin our data layers based on the assumption that new forestregrowth would not meet the criteria for intact forestwithin this time period. The annual products were pro-duced sequentially, with results from the preceding yearused to create those for the subsequent year, thus pro-ducing output that tracked the decline in forest coverand any areas of regrowth. Once produced, the 500 mcategorical results for each year were converted to frac-tional (i.e. continuous) products at a 5 km resolution, withvalues ranging from 0.0 (no forest cover) to 1.0 (completeforest cover) for the proportion of each 5 × 5 km classifiedas each forest type. Further details on the constructionand validation of the forest covariate data surfaces are pro-vided in Additional file 2.ModelThe boosted regression trees method used here is avariant of the model used in previous analyses of malariavector species [24] and diseases such as dengue [28].Boosted regression tree modelling combines both regres-sion trees (which build a set of decision rules on the pre-dictor variables by portioning the data into successivelysmaller groups with binary splits) and boosting (whichselects sets of trees that minimise the loss of function)to best capture the variables that define the distributionof the input data [29–31]. The core setup used in thecurrent study has been described previously [28]. Thechanges made to the method for this study allowedtemporal changes in land cover to be incorporated andimproved the way absence data and sampling bias werehandled. Our methods for handling polygon data alsovaried from those used previously.The previous approaches [24, 28] used synoptic covariatevalues that covered a period of several years. In this study,we incorporated temporal covariate data for the land coverclasses so that the year of occurrence was taken into ac-count. We were able to construct covariate data layers foreach year from 2001 to 2012 but the species data availablefor this period did not cover all of the geographical regionsfor which we had data. To test the impact of using speciesdata that could not be matched to the corresponding an-nual land cover layers we constructed two test datasets forM. leonina and the Dirus Complex for the time period2001 to 2012 only. We ran the model as described belowtwice on each dataset; once linking the species data to an-nual covariate layers and once linking the species data tothe 2012 covariate layers only. The results provided inAdditional file 3 show that the outputs from the two modelversions were similar but the version using annual covari-ates performed slightly better. We therefore used the fulldataset for the final model runs, including data outside the2001 to 2012 year range, in order to maximise the spreadof species data used, and we linked species data from 2001to 2012 to covariate values for the relevant year in order toimprove the predictions where possible. For all occurrencesprior to 2001, covariate values for 2001 were used, and forany data collected after 2012, covariate values for 2012were used. Model predictions were made to the most con-temporary covariate data available.The boosted regression trees method requires bothpresence and absence data. Pseudo-absence data, alsoknown as background data, are generated when true ab-sence data is not available. The vast majority of speciesoccurrence datasets are subject to spatial bias, for ex-ample, areas near roads and paths may be more likely tobe surveyed than other sites. If unaccounted for, this sur-vey bias can translate into environmental bias in the fittedmodel. One approach for coping with biased occurrencedata is to select background data that reflect the samesurvey bias as the occurrence data. The resulting modelshould identify suitable environments for the specieswithin the sampled space, rather than just areas that aremore heavily sampled. This approach does not eliminatesampling bias issues entirely but improved model per-formance has been demonstrated when compared to theuse of randomly selected background data [32]. For thework presented here, mammal and malaria vector occur-rence records from within the study area were used aspseudo-absence data for the macaque and vector models,respectively. These background datasets were chosen be-cause the sampling methods were the same as those usedfor the target species.For each species, we fitted 120 submodels each trainedto a randomly selected bootstrap of the presence/back-ground dataset. Each bootstrap contained a minimum offive presence and five background points. To accountfor uncertainty in the geographic location of occurrenceslinked to polygon locations > 25 km2 in the M. leoninadataset, one 5 × 5 km pixel within each polygon wasrandomly selected for each bootstrap. Each of the sub-models generated a predicted value for the relativeprobability of species occurrence at every 5 × 5 km pixeland together the submodels generated a distribution ofMoyes et al. Parasites & Vectors  (2016) 9:242 Page 4 of 12predicted values for every pixel. We generated maps dis-playing the mean, 0.025 quantile and 0.975 quantilevalues from these distributions for each pixel.To evaluate the ensemble’s predictive performance, wecalculated, for each submodel, the area under the receiveroperator curve (AUC), i.e. the area under a plot of the truepositive rate versus the false positive rate, reflecting theability to discriminate between presence and backgroundrecords, whilst marginalising the arbitrary choice of a clas-sification threshold [33]. For each submodel, we reportedthe mean AUC under fivefold cross-validation using apairwise distance sampling procedure to remove spatialsorting bias in the model validation [34]. We then com-bined these submodel validation statistics to obtain anoverall estimate of predictive performance in the ensem-ble, and uncertainty in this estimate.There were insufficient presence data for An. cracens,An. balabacensis, An. latens, or An. introlatus to modelthese species individually so members of the Dirus andLeucosphyrus Complexes were modelled collectively topredict the relative probability of one or more of thespecies within a complex occurring.Covariate density plotsTo illustrate the relationships between the coverage ofeach land cover class and predicted species occurrence,we plotted the relative density of pixels at each percentagecoverage for all pixels where the relative probability ofspecies occurrence was greater than 0.75. This thresholdwas selected in order to identify pixels with a high prob-ability of occurrence rather than simply those with a rela-tive probability greater than 50 %. The density values werecalculated from the ratio of the number of pixels wherethe relative probability of species occurrence was greaterthan 0.75 to the total number of pixels in the study area.MasksThe model outputs for each species were restricted tothe islands within its known range, using the range mapsdeveloped for this project. For the macaque species,range maps were obtained from the International Unionfor the Conservation of Nature [35]. These ranges didnot incorporate all new data or introduced populations.We therefore used our occurrence dataset to adjust therange for each species by either dragging the range outto encompass new reports next to the existing range orby identifying the borders of any confirmed populationthat was not contiguous with the existing range. For thisexercise, data did not need to meet the criteria of repre-senting an area < 25 km2. For the mosquito species/complexes, we used range maps previously published bythree groups [24, 36, 37] and updated these in the sameway as the macaque ranges. In places where the threeranges differed for a particular species, we selected thebroadest of the available options.ResultsMacaque and mosquito distributionsThe mean model outputs, masked out on islands out-side the species range, were used to generate predictedspecies distribution maps (Figs. 1 and 2). The AUCvalues ± standard error for the predicted macaques dis-tributions were 0.858 ± 0.001 for the M. fascicularismap, 0.821 ± 0.003 for the M. nemestrina map and 0.830 ±0.002 for the M. leonina map. The AUC values ± standarderror for the predicted mosquito distributions were 0.860 ±0.005 for the An. dirus map, 0.885 ± 0.002 for the DirusComplex map, 0.842 ± 0.009 for the Leucophyrus Complexmap and 0.883 ± 0.001 for the Leucophyrus Group map.The Leucosphyrus Complex predictions should be inter-preted with caution because the data volume for thisComplex was low and the data was sparse. The model didnot predict many areas with a high probability of occur-rence outside the current macaque ranges, indicating thateach species has largely realised its predicted niche,excluding islands that have not yet been populated(Fig. 1).The 0.025 and 0.975 quantile from the model ensemble,and the top predictors for each species together withvalues for their relative influence on the model, are pro-vided in Additional files 4 and 5.Density plots illustrating the ratio of proportional landcover in areas with high predicted probability of occur-rence (predictions of 0.75 and above) to proportionalland cover in all areas are given in Additional file 6. If allland cover proportions (for example high, low and nocoverage of grassland) were equally likely to be suitable(a null distribution), each of these density plots would beflat. High or low regions of the density plots thereforeindicate proportional land cover values which are pre-dicted to be more or less suitable for occurrence of thespecies in question. These plots should not be inter-preted as providing robust evidence for any specific rela-tionship between each land cover class and speciespresence, in part because each plot is influenced by therelationships between species occurrence and all of theother covariates that went into the model for the locationswhere we had data. The purpose of these plots is solely toprovide an extra visualisation of the predicted species dis-tribution results and supplement the information that canbe visualised directly on the maps in Figs. 1 and 2.Data availabilityAll data used and generated by the project are publiclyavailable. The vector occurrence data used in our modelsare available from http://www.map.ox.ac.uk/explorer/#EntityPlace:Anopheline. The macaque occurrence datasets usedMoyes et al. Parasites & Vectors  (2016) 9:242 Page 5 of 12in each model are provided in Additional files 7, 8 and 9.The GeoTIFF files containing mean model outputs as dis-played in the distribution maps are provided in Additionalfiles 10, 11, 12, 13, 14, 15 and 16. The model code gener-ated is available on an open source basis from https://github.com/SEEG-Oxford/seegSDM.DiscussionForest cover is shrinking in southeast Asia and the lossof natural or intact forest is in part due to conversion todegraded or logged areas and plantations (included inthe disturbed forest category in this study) [38]. Further-more, this trend is expected to continue [39]. This studyis the first to predict the full distributions of knowlesimalaria host and vector species by modelling these spe-cies across their ranges using environmental data sur-faces that track changes in land cover. The predicteddistributions generated are not solely restricted to for-ested areas and any disease risk in non-forested areashas the potential to dramatically increase estimates ofthe population at potential risk of knowlesi malariawithin this densely populated region [5].The species distributions were generated using envir-onmental covariates but the relationship between speciesoccurrence and these covariates may vary in time andspace. One advantage of the BRT approach is its abilityto fit a single overall model to multiple distinct patternswithin the data. This flexible nonparametric statisticalmodel is better able to simultaneously model multipleenvironmental relationships than more traditional ap-proaches [29]. The cost of this flexibility and focus onpredictive power is that it complicates inference, thus, itis not possible to identify causal relationships betweenthe environmental covariates and the species modelled.For the vector species, data scarcity meant we had topool data from multiple species and again BRT’s abilityto encompass different relationships within a pooleddataset helps to overcome some of the issues of model-ling multiple species together. The model is still limitedas it is only able to model relationships for the combina-tions of covariates found within the field data providedto the model.The model is also limited by the set of covariates used.These do not capture all potential sources of variationthat may influence the distributions of these species insome or all parts of their ranges. For example, forestedge effects were not incorporated in our study. Previ-ously the booted macaque, M. ochreata, has been shownto be more abundant near forest edges at two sites inSulawesi [40], however, a study of tiger prey, includingFig. 1 Ranges and predicted distributions of the macaque species. The three maps on the left show the current range of each macaque speciesand the three maps on the right show the predicted relative probability of occurrence at every 5 × 5 km pixel within the study area on a scale of0 to 1.0Moyes et al. Parasites & Vectors  (2016) 9:242 Page 6 of 12M. nemestrina, on Sumatra found that edge effects asso-ciated with national park boundaries were not significantonce human population density was considered [41].The potential predictors provided to the model arealso limited by the fact that each land class encompassesvariation in that habitat; most noticeably the disturbedforest class includes natural forest with evidence of humandisturbance, or less than 500 km2 in area, and establishedpalm and rubber plantations. Our aim was to model eachspecies across its entire range using region-wide covariatedatasets rather than more detailed but locally-restricteddata. Furthermore, our approach did not incorporate fac-tors that may be important at finer resolutions than the5 km used here [42]. Nevertheless, the models performedreasonably well using the covariate data that we con-structed for the region as a whole, at a 5 km resolution,and gave AUC values of 0.82 and above. The answers tomore detailed questions about the influence of individualfactors on host and vector populations require moredetailed studies. The distributions presented here providea good estimate of the full distributions of the species im-portant to P. knowlesi transmission across their ranges.Earlier studies of M. nemestrina were more detailedbut also more geographically restricted. In IndonesianBorneo, M. nemestrina was found in intact and partiallydegraded or burned forest but was absent from completelydeforested areas [43]. On Sumatra, M. nemestrina densitieswere higher in areas of low human population density [41]and this species was less common in plantations comparedto M. fascicularis, although it was found [44]. Macacanemestrina raids rice crops in Sumatra, most frequently onfarms close to the forest edge [45]. Of the macaque speciesstudied here, our model predicted a high relative probabilityof M. nemestrina presence in the smallest number ofland cover classes, and rarely predicted occurrence innon-forested areas. This finding and those of otherstudies support the hypothesis that this species will benegatively impacted by conversion of forested areas toFig. 2 Ranges and predicted distributions of the mosquito species, complexes and group. The four maps on the left show the current range of eachmosquito species, complex or group, and the four maps on the right show the predicted relative probability of occurrence at every 5 × 5 km pixelwithin the study area on a scale of 0 to 1.0Moyes et al. Parasites & Vectors  (2016) 9:242 Page 7 of 12non-forest habitats but conversion of intact forest todisturbed forest will allow populations to remain. Thelatter conversion will, however, bring humans into thevicinity of these macaque populations where they hadpreviously been separated.Our model predicted that the M. fascicularis and M.leonina distributions encompass a wider range of habi-tats than the M. nemestrina distribution, and this is par-ticularly apparent for the M. fascicularis distribution. Anumber of studies have measured the habitat preferencesof M. fascicularis within restricted parts of its range butnone have considered the full distribution of this species.In Malaysian Borneo, studies found that M. fascicularispopulations were initially negatively impacted by loggingactivities but their local abundance was higher in areasthat had been logged ten years previously than inunlogged forest [46]. In Vietnam, M. fascicularis wasfound in public parks and temples as well as mangroves,river banks and primary, disturbed and secondary forests[47]. On Sumatra, M. fascicularis was found in plantationsmore commonly than M. nemestrina [44]. In Thailand, M.fascicularis habitats have changed from natural forests totemples and parks over the last 30 years [48] and groupsare found in suburban Bangkok, the Thai capital [49]. Thisspecies will also freely enter suburban areas of SelangorState, Malaysia [50]. In Singapore, M. fascicularis is foundat forest perimeters and in forest fragments, where thesemacaques are habituated to human presence and willleave forest areas for urban habitats [51]. Our results andthese previous findings strongly suggest that M. fascicu-laris populations are able to occupy a wide range ofhabitats. Importantly, the distribution of this speciesencompasses many locations close to human habitation(urban areas) or activity (disturbed forests, orchards,croplands, etc).There are fewer studies of M. leonina habitat prefer-ences. In recent surveys, M. leonina monkeys were foundin a range of habitats in Laos from river areas to inter-mediate plains to dry hilly forest [52]. When the habitatpreference of this species was measured within a Laos re-serve, it was associated with proximity to village areas(average 6 km) as well as with evergreen and deciduousforest cover, lower elevations and higher temperatures[53]. Macaca leonina has also been found to move be-tween primary and secondary forest in a Thai reserve [54].The evidence available, and our own results, suggests thatthis species occurs in deforested areas where human ac-tivity occurs although, like M. nemestrina, this specieswas not associated with urban areas.The question of whether these latter two macaquespecies occur in northern Myanmar is of particular rele-vance because human cases of P. knowlesi malaria havebeen found in people living in Shan State near theborder with Yunnan Province, China [16, 17]. If thecurrent ranges used here are accurate then, of the specieswe mapped for this study, only M. leonina is present innortheast Myanmar. Records of macaques in Myanmarare, however, incomplete and may be out of date [55]although a recent survey close to the Yunnan border inKachin State (to the north of Shan State) found M. leoninamonkeys but not M. fascicularis [56]. The M. fascicularisrange extends at least as far north as central Myanmar butour results predict that the habitats further northeast areunsuitable for this species (Fig. 1). Further surveys of thisregion are necessary to confirm the full list of speciespresent and whether they are infected with P. knowlesi.Sulawesi is not in the natural range of the three ma-caque species studied here although related species arefound on the island [57]. Macaca fascicularis and M.nemestrina monkeys are kept as pets on Sulawesi [58]and there is an unsubstantiated report of P. knowlesiinfecting a macaque here [59]. For these reasons we haveshown the model predictions for these species on Sulawesibut this area of the map should be interpreted as showingsuitable environments for these species should they estab-lish feral populations.The Leucosphyrus Complex of mosquitoes was pre-dicted to occur in areas with high coverage of disturbedforest but lacking intact forest cover, although the sparsedata and low data volumes for this Complex mean thesepredictions need to be interpreted with caution. ThisComplex is responsible for P. knowlesi transmission inthe region where knowlesi malaria is believed to be mostcommon, Malaysian Borneo, and where deforestation(the loss of intact forest) is occurring [21, 60]. Notifica-tions of knowlesi malaria cases increased in MalaysianBorneo between 1992 and 2011 [61] and our results in-dicate the conversion of intact forest to disturbed forest,and the resulting impact on the probability of encoun-tering members of the Leucosphyrus Complex, could bea factor here. One study in the northern part of Sabah inMalaysian Borneo recently found an association betweentwo forest variables, forest loss and total cover within2 km of a village, and the estimated incidence of knowlesimalaria at the village level in the two districts studied butthey did not distinguish intact and disturbed forest [62].The small number of previous studies of the LeucosphyrusComplex all focussed on measuring the characteristics ofrelevance to vectorial capacity rather than relationshipswith environmental factors. These studies were conductedin restricted geographical areas and did not explicitlymeasure environmental variables, although sites wereclassified into types. Anopheles balabacensis in an area ofSabah was more abundant in a village site than the forestsite and farming site surveyed, but P. knowlesi infectionrates were lowest in the village [21]. Anopheles latenshuman biting rates in Sarawak, Malaysia were higher ata fruit tree farm on the forest fringe and a forest siteMoyes et al. Parasites & Vectors  (2016) 9:242 Page 8 of 12compared to a longhouse site, and P. knowlesi sporozo-ite infected individuals were only found at the fruit treeand forest sites [60]. Ours is the first study to investi-gate the distribution of the Leucosphyrus Complex,generally considered to be a forest species [24, 36, 37]and to have predicted occurrence in areas with dis-turbed forest cover.Two previous studies have considered the relation-ships between the Dirus Complex and environmentalfactors. The research presented here builds on an earlierproject that predicted distributions for all primary hu-man malaria vectors in the region using an older versionof the field dataset used here, and similar methods [24].In the current study we developed the modelling tech-niques further to handle sampling bias, we extended therange of covariates tested and we incorporated data ontemporal changes to land cover since 2001. We foundsimilar relationships with environmental factors to theearlier work and observed a closer match to the knownrange of the Dirus Complex. An independent study useda different approach to model both the potential nicheand the current distribution or realised niche of thisComplex [63]. They used temperature and rainfall to de-fine the fundamental niche, and land cover (specificallyforest cover in 2005) to model the realised niche orcurrent distribution. Our results did not predict DirusComplex occurrence exclusively in forested areas butthe outputs from the two studies, our predicted distribu-tion and their realised niche map, show similar results.More detailed but geographically restricted studies haveconsidered differences in abundance in different habitattypes at a small number of sites but have not explicitlyanalysed the relationships with environmental variables.The An. dirus biting and infection rates did not differsignificantly between a forest site and a forest fringe sitein Khanh Hoa Province, Vietnam [64], whereas An. cracenswas more abundant and had a higher human biting rate atan orchard site compared to a forest edge and village sitein Pahang State, Malaysia [19]. Our study took data from amuch larger number of sites and used quantified environ-mental variables, but is restricted to species occurrence.For all potential vector species, and particularly themembers of the Leucosphyrus Complex implicated as P.knowlesi vectors, more data is needed to define their dis-tributions with confidence. Large-scale systematic sam-pling of a range of habitats across the region is urgentlyneeded to address this important gap in our understandingof P. knowlesi infection risk.ConclusionsTogether our results for the macaque hosts and the mos-quito vectors of P. knowlesi malaria suggest that the rela-tive probability of host macaque species and members ofthe Leucosphyrus Complex occurring in disturbed forestareas, for example, plantations or timber concessions, andvegetation mosaics, will mean these species can co-existclose to human activity. This finding is of most signifi-cance in the southern part of our study area (Malaysia,Indonesia, Singapore, Brunei and part of the Philippines)where members of this Complex are the main P. knowlesivectors. The predicted distribution of the long-tailed ma-caque, M. fascicularis, encompassed many more types ofhuman-occupied habitat. This is of most significance in thenorthern part of its range and our study area (Myanmar,Thailand, Laos, Cambodia and Vietnam) because membersof the Dirus Complex are the main P. knowlesi vectors hereand our model predicts occurrence of these species inareas of open canopy cover (savannah), vegetation mosaicsand cropland as well as closed canopy forest.Characterising the distribution of all component speciesis an important first step in understanding the distributionof a vector-borne, zoonotic disease when human infectiondata is lacking. The maps generated here will help identifyareas where there may be a P. knowlesi disease risk butfurther information is needed to extrapolate directly fromthese maps to an index of risk [42]. The next stage of thiswork, therefore, needs to consider the relationship be-tween P. knowlesi infections and a range of risk factorsincluding the fine scale species distributions presentedhere as well as geographical, environmental and socioeco-nomic factors. Using a similar modelling framework, theP. knowlesi reservoir and vector maps can be used as ex-planatory variables to test their ability to predict spatialvariation in risk of human P. knowlesi infections in areaswhere human disease data is available. The resultingmodel could then be used to predict human disease riskin areas where both reservoirs and competent vectors arelikely to be present but human disease data is scarce orabsent. Only then can we consider estimating the popula-tion at risk across the region.Additional filesAdditional file 1: Distributions of the model input data in space andtime. The spatial distributions of the species occurrence data andbackground data are shown on a series of maps and their temporaldistributions are shown by a series of histograms. (DOCX 1584 kb)Additional file 2: Environmental variables used in the species distributionmodels and construction of the forest cover layers. Details of the covariatedata used are provided together with their source details. The constructionand validation of the intact and disturbed forest layers is described in detail.(DOCX 27 kb)Additional file 3: Investigating the impact of using annual land cover data.The model was run using identical datasets and either 1) year-matched landcover data or 2) 2012 land cover data. The resulting distributions are shownwith the AUC ± standard error, and the top predictors with their relativeinfluences. A map showing the difference between the two resultingdistributions is also provided. (DOCX 2977 kb)Additional file 4: The 0.025 and 0.975 quantile model predictions, andthe top predictors, for each macaque species. For each macaque speciesthe 0.025 and 0.975 quantile model outputs, masked out on islands outsideMoyes et al. Parasites & Vectors  (2016) 9:242 Page 9 of 12each species range, are provided with the mean AUC (± standard error) andthe relative influence of the top predictors for that species. (DOCX 1183 kb)Additional file 5: The 0.025 and 0.975 quantile model predictions, andthe top predictors, for each mosquito model. For each mosquito species,complex or group, the 0.025 and 0.975 quantile model outputs, maskedout on islands outside each species or complex range, are provided withthe mean AUC (± standard error) and the relative influence of the toppredictors for that model. (DOCX 1331 kb)Additional file 6: Proportional land cover in areas with high predictedprobability of species occurrence. Plots showing the relative density of pixelsat each percentage land class coverage for all pixels where the probability ofspecies occurrence was greater than 0.75, for each species. (DOCX 795 kb)Additional file 7: Macaca fascicularis data. Each record of M. fascicularisoccurrence is provided with a location and date. Duplicate records withina calendar year have been removed. Locations are classed as points(defined as <25 km2) or polygons (defined as >25 km2). (XLSX 322 kb)Additional file 8: Macaca nemestrina data. Each record of M. nemestrinaoccurrence is provided with a location and date. Duplicate records withina calendar year have been removed. Locations are classed as points(defined as <25 km2) or polygons (defined as >25 km2). (XLSX 130 kb)Additional file 9: Macaca leonina data. Each record of M. leoninaoccurrence is provided with a location and date. Duplicate records withina calendar year have been removed. Locations are classed as points(defined as <25 km2) or polygons (defined as >25 km2). (XLSX 63 kb)Additional file 10: Relative probability of Macaca fascicularis occurrence.A GeoTIFF raster data layer containing a predicted value (the meanmodel output) for every 5×5km pixel within SE Asia excluding islandsoutside the species range. This file can be opened in GIS software (e.g.QGIS, ArcMap, etc) or using the ‘raster’ R package. (TIF 4459 kb)Additional file 11: Relative probability of Macaca nemestrina occurrence.A GeoTIFF raster data layer containing a predicted value (the mean modeloutput) for every 5×5km pixel within SE Asia excluding islands outside thespecies range. This file can be opened in GIS software (e.g. QGIS, ArcMap,etc) or using the ‘raster’ R package. (TIF 4090 kb)Additional file 12: Relative probability of Macaca leonina occurrence. AGeoTIFF raster data layer containing a predicted value (the mean modeloutput) for every 5×5km pixel within SE Asia excluding islands outsidethe species range. This file can be opened in GIS software (e.g. QGIS,ArcMap, etc) or using the ‘raster’ R package. (TIF 3459 kb)Additional file 13: Relative probability of Anopheles dirus occurrence. AGeoTIFF raster data layer containing a predicted value (the mean modeloutput) for every 5×5km pixel within SE Asia excluding islands outsidethe species range. This file can be opened in GIS software (e.g. QGIS,ArcMap, etc) or using the ‘raster’ R package. (TIF 3461 kb)Additional file 14: Relative probability of a member of the Dirus Complexoccurring. A GeoTIFF raster data layer containing a predicted value (the meanmodel output) for every 5×5km pixel within SE Asia excluding islands outsidethe complex range. This file can be opened in GIS software (e.g. QGIS, ArcMap,etc) or using the ‘raster’ R package. (TIF 3770 kb)Additional file 15: Relative probability of a member of the LeucosphyrusComplex occurring. A GeoTIFF raster data layer containing a predicted value(the mean model output) for every 5×5km pixel within SE Asia excludingislands outside the complex range. This file can be opened in GIS software(e.g. QGIS, ArcMap, etc) or using the ‘raster’ R package. (TIF 4006 kb)Additional file 16: Relative probability of a member of the LeucosphyrusGroup occurring. A GeoTIFF raster data layer containing a predicted value(the mean model output) for every 5×5km pixel within SE Asia excludingislands outside the group range. This file can be opened in GIS software(e.g. QGIS, ArcMap, etc) or using the ‘raster’ R package. (TIF 4524 kb)AbbreviationsAUC: area under the receiver operator curve; BRT: boosted regression tree;IFL: intact forest landscape; IGBP: International Geosphere and BiosphereProgramme.Competing interestsThe authors declare that they have no competing interests.Authors’ contributionsCLM, FS and NG designed the overarching study. ZH, DW and CLM designedand constructed the forest cover data layers. HG constructed the otherenvironmental data layers with input from DW. FS developed the modelswith advice from NG, DMP and CLM. VN, JM-A, JFB, SM, ML, HS, TGO, CRT,YH and AJG advised on the macaque modelling and contributed data. IV,MES, IRFE and MJB advised on the vector modelling and ranges. ZH andCLM constructed the species ranges, and AW and CLM collated, standardisedand geopositioned the occurrence data. All authors contributed to theinterpretation of the results presented here and approved the final version ofthe manuscript.AcknowledgementsThe authors wish to thank Thomas Brooks (IUCN), Camille Coudrat,Herbert H. Covert (University of Colorado Boulder), Hoang Minh Duc (SouthernInstitute of Ecology), Antje Engelhardt, Gabriella Fredriksson, Juan-CarlosGonzalez (Institute of Biological Sciences), Tom Gray (WWF-Cambodia, and theForestry Administration and Ministry of the Environment of the RoyalCambodian Government), Islamul Hadi (Mataram University), Kate Jenks(Smithsonian Conservation Biology Institute), Dede Aulia Rahman(Bogor Agricultural University), Rustam (Faculty of Forestry, MulawarmanUniversity), Jonathan O’Brien (University of Colorado Boulder) and AchmadYanuar (Universitas Nasional Jalan) for providing unpublished and pre-publicationdata and coordinates. We are grateful to the Sarawak Forestry Corporation andForest Department Sarawak who provided research permits to JayasilanMohd-Azlan.None of the funders had any role in the design, collection, analysis, orinterpretation of data; or in the writing of the manuscript; or in the decisionto submit the manuscript for publication.Author details1Spatial Ecology & Epidemiology Group, The Big Data Institute, Li Ka ShingCentre for Health Information and Discovery, University of Oxford, OxfordOX3 7BN, UK. 2Spatial Ecology & Epidemiology Group, Department ofZoology, University of Oxford, Oxford OX1 3PS, UK. 3Department of SocialSciences, Oxford Brookes University, Oxford OX1 0BP, UK. 4Department ofZoology, Faculty of Resource Science and Technology, Universiti MalaysiaSarawak, 94300 Kota Samarahan, Sarawak, Malaysia. 5Departments of Zoologyand Botany, University of British Columbia, Vancouver, BC V6T 1Z4, Canada.6Biodiversity Research Centre, University of British Columbia, Vancouver, BCV6T 1Z4, Canada. 7Primate Research Unit, Department of Biology, Faculty ofScience, Chulalongkorn University, Bangkok 10330, Thailand. 8Fauna & FloraInternational, Singapore 247672, Singapore. 9Institute for GlobalEnvironmental Strategies, Kamiyamaguchi 2108-11, Hayama-cho 240-0115,Kanagawa, Japan. 10Wildlife Conservation Society, Mpala Research Center,Nanyuki 10400, Kenya. 11Research Institute for the Environment andLivelihoods, Charles Darwin University, Northern Territory 0909, Australia.12Faculty of Science and Technology, Federation University Australia, MtHelen, Victoria 3350, Australia. 13Evolutionary Morphology Section, PrimateResearch Institute, Kyoto University, Inuyama, Japan. 14Field ConservationProgram, S.P.E.C.I.E.S., Ventura, CA, USA. 15Conservation Science Program,Tiger Creek Wildlife Refuge, Tyler, TX, USA. 16Eijkman-Oxford Clinical ResearchUnit, Jakarta, Indonesia. 17Department of Parasitology, Faculty of Medicine,University of Malaya, Kuala Lumpur, Malaysia. 18Department of Entomology,Faculty of Agriculture, Kasetsart University, Bangkok 10900, Thailand. 19PublicHealth and Malaria Control Department, International SOS, Jalan Kertajasa,Kuala Kencana, Papua 99920, Indonesia. 20Institute for Health Metrics andEvaluation, University of Washington, Seattle, WA 98121, USA. 21WellcomeTrust Centre for Human Genetics, University of Oxford, Oxford OX3 7BN, UK.Received: 9 February 2016 Accepted: 21 April 2016References1. World Health Organization. World malaria report 2014. Geneva: WorldHealth Organization; 2014.2. Barber BE, William T, Grigg MJ, Menon J, Auburn S, Marfurt J, Anstey NM,Yeo TW. A prospective comparative study of knowlesi, falciparum, and vivaxMoyes et al. Parasites & Vectors  (2016) 9:242 Page 10 of 12malaria in Sabah, Malaysia: High proportion with severe disease fromPlasmodium knowlesi and Plasmodium vivax but no mortality with earlyreferral and artesunate therapy. Clin Infect Dis. 2013;56(3):383–97.3. Singh B, Daneshvar C. Human infections and detection of Plasmodiumknowlesi. Clin Microbiol Rev. 2013;26(2):165–84.4. Moyes CL, Henry AJ, Golding N, Huang Z, Singh B, Baird JK, Newton PN,Huffman M, Duda KA, Drakeley CJ, et al. Defining the geographical range ofthe Plasmodium knowlesi reservoir. PLoS Negl Trop Dis. 2014;8(3):e2780.5. WHO Regional Office for the Western Pacific. Informal consultation on thepublic health importance of Plasmodium knowlesi. Manila: World HealthOrganization; 2011. p. 36.6. Cox-Singh J. Knowlesi malaria in Vietnam. Malar J. 2009;8:269.7. Millar SB, Cox-Singh J. Human infections with Plasmodium knowlesi-zoonoticmalaria. Clin Microbiol Infect. 2015;21(7):640–8.8. Divis PCS, Singh B, Anderios F, Hisam S, Matusop A, Kocken CH, Assefa SA,Duffy CW, Conway DJ. Admixture in humans of two divergent Plasmodiumknowlesi populations associated with different macaque host species. PLoSPath. 2015;11(5):e1004888.9. Cox-Singh J, Culleton R. Plasmodium knowlesi: from severe zoonosis toanimal model. Trends Parasitol. 2015;31(6):232–8.10. Mueller M, Schlagenhauf P. Plasmodium knowlesi in travellers, update 2014.Int J Infect Dis. 2014;22:55–64.11. Jongwutiwes S, Buppan P, Kosuvin R, Seethamchai S, Pattanawong U,Sirichaisinthop J, Putaporntip C. Plasmodium knowlesi malaria in humansand macaques, Thailand. Emerg Infect Dis. 2011;17(10):1799–806.12. Lee KS, Divis PC, Zakaria SK, Matusop A, Julin RA, Conway DJ, Cox-Singh J,Singh B. Plasmodium knowlesi: reservoir hosts and tracking the emergencein humans and macaques. PLoS Pathog. 2011;7(4):e1002015.13. Assefa S, Lim C, Preston MD, Duffy CW, Nair MB, Adroub SA, Kadir KA,Goldberg JM, Neafsey DE, Divis P, et al. Population genomic structure andadaptation in the zoonotic malaria parasite Plasmodium knowlesi. Proc NatlAcad Sci U S A. 2015;112(42):13027–32.14. Roos C, Thanh VN, Walter L, Nadler T. Molecular systematics of Indochineseprimates. Vietn J Primatol. 2007;1(1):41–53.15. Malaivijitnond S, Arsaithamkul V, Tanaka H, Pomchote P, Jaroenporn S,Suryobroto B, Hamada Y. Boundary zone between northern and southernpig-tailed macaques and their morphological differences. Primates. 2012;53(4):377–89.16. Zhu H, Li J, Zheng H. Human natural infection of Plasmodium knowlesi.Chin J Parasitol Parasit Dis. 2006;24(1):70.17. Jiang N, Chang Q, Sun X, Lu H, Yin J, Zhang Z, Wahlgren M, Chen Q. Co-infections with Plasmodium knowlesi and other malaria parasites, Myanmar.Emerg Infect Dis. 2010;16(9):1476–8.18. Marchand RP, Culleton R, Maeno Y, Quang NT, Nakazawa S. Co-infections ofPlasmodium knowlesi, P. falciparum, and P. vivax among humans and Anophelesdirus mosquitoes, Southern Vietnam. Emerg Infect Dis. 2011;17(7):1232–9.19. Jiram AI, Vythilingam I, NoorAzian YM, Yusof YM, Azahari AH, Fong M-Y.Entomologic investigation of Plasmodium knowlesi vectors in Kuala Lipis,Pahang, Malaysia. Malar J. 2012;11:213.20. Vythilingam I, Tan CH, Asmad M, Chan ST, Lee KS, Singh B. Naturaltransmission of Plasmodium knowlesi to humans by Anopheles latens inSarawak, Malaysia. Trans R Soc Trop Med Hyg. 2006;100(11):1087–8.21. Wong M, Chua T, Leong C, Khaw LT, Fornace K, Wan-Sulaiman W-Y, William T,Drakeley CJ, Ferguson H, Vythilingam I. Seasonal and spatial dynamics of theprimary vector of Plasmodium knowlesi within a major transmission focus inSabah, Malaysia. PLoS Negl Trop Dis. 2015;9(10):e0004135.22. Vythilingam I, Lim YAL, Venugopalan B, Ngui R, Leong CS, Wong ML, Khaw L,Goh X, Yap N, Sulaiman WYW, et al. Plasmodium knowlesi malaria an emergingpublic health problem in Hulu Selangor, Selangor, Malaysia (2009-2013):epidemiologic and entomologic analysis. Parasit Vectors. 2014;7:436.23. Wharton RH, Eyles DE, Cheong WH, Warren M. Studies to determine vectorsof monkey malaria in Malaya. Ann Trop Med Parasitol. 1964;58(1):56–77.24. Sinka ME, Bangs MJ, Manguin S, Chareonviriyaphap T, Patil AP, TemperleyWH, Gething PW, Elyazar IRF, Kabaria CW, Harbach RE, et al. The dominantAnopheles vectors of human malaria in the Asia-Pacific region: occurrencedata, distribution maps and bionomic precis. Parasit Vectors. 2011;4:89.25. Weiss DJ, Mappin B, Dalrymple U, Bhatt S, Cameron E, Hay SI, Gething PW.Re-examining environmental correlates of Plasmodium falciparum malariaendemicity: a data-intensive variable selection approach. Malar J. 2015;14(1):574.26. User Guide for the MODIS Land Cover Type Product (MCD12Q1) [http://www.bu.edu/lcsc/files/2012/08/MCD12Q1_user_guide.pdf]. Accessed 20 Apr 2014.27. Potapov P, Yaroshenko A, Turubanova S, Dubinin M, Laestadius L, Thies C,Aksenov D, Egorov A, Yesipova Y, Glushkov I, et al. Mapping the world’sintact forest landscapes by remote sensing. Ecol Soc 2008; 13(2):51.28. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL, Drake JM,Brownstein JS, Hoen AG, Sankoh O, et al. The global distribution andburden of dengue. Nature. 2013;496(7446):504–7.29. Elith J, Leathwick JR, Hastie T. A working guide to boosted regression trees.J Anim Ecol. 2008;77(4):802–13.30. De’ath G. Boosted trees for ecological modeling and prediction. Ecology.2007;88(1):243–51.31. Elith J, Graham CH, Anderson RP, Dudik M, Ferrier S, Guisan A, Hijmans RJ,Huettmann F, Leathwick JR, Lehmann A, et al. Novel methods improveprediction of species’ distributions from occurrence data. Ecography. 2006;29(2):129–51.32. Phillips S, Dudik M, Elith J, Graham C, Lehmann A, Leathwick J, Ferrier S.Sample selection bias and presence-only distribution models: implicationsfor background and pseudo-absence data. Ecol Appl. 2009;19:181–97.33. Fleiss J, Levin B, Paik M. Statistical methods for rates and proportions.Hoboken: John Wiley & Sons; 2003.34. Hijmans RJ. Cross-validation of species distribution models: removing spatialsorting bias and calibration with a null model. Ecology. 2012;93(3):679–88.35. Macaca range maps. In: IUCN Red List of Threatened Species. Version 2013.2. [http://www.iucnredlist.org/]. Accessed 12 Jun 2013.36. Manguin S, Garros C, Dusfour I, Harbach RE, Coosemans M. Bionomics,taxonomy, and distribution of the major malaria vector taxa of Anophelessubgenus Cellia in Southeast Asia: An updated review. Infect Genet Evol.2008;8(4):489–503.37. Sallum MAM, Peyton EL, Harrison BA, Wilkerson RC. Revision of theLeucosphyrus group of Anopheles (Cellia) (Diptera, Culicidae). Rev BrasEntomol. 2005;49:1–152.38. Hansen MC, Potapov PV, Moore R, Hancher M, Turubanova SA, Tyukavina A,Thau D, Stehman SV, Goetz SJ, Loveland TR, et al. High-resolution globalmaps of 21st-century forest cover change. Science. 2013;342(6160):850–3.39. Carlson KM, Curran LM, Ratnasari D, Pittman AM, Soares-Filho BS, Asner GP,Trigg SN, Gaveau DA, Lawrence D, Rodrigues HO. Committed carbonemissions, deforestation, and community land conversion from oil palmplantation expansion in West Kalimantan, Indonesia. Proc Natl Acad Sci U SA. 2012;109(19):7559–64.40. Froese GZL, Contasti AL, Mustari AH, Brodie JF. Disturbance impacts onlarge rain-forest vertebrates differ with edge type and regional context inSulawesi, Indonesia. J Trop Ecol. 2015;31:509–17.41. O’Brien TG, Kinnaird MF, Wibisono HT. Crouching tigers, hidden prey:Sumatran tiger and prey populations in a tropical forest landscape. AnimConserv. 2003;6(2):131–9.42. Brock PM, Fornace KM, Parmiter M, Cox J, Drakeley CJ, Ferguson HM, Kao RR.Plasmodium knowlesi transmission: integrating quantitative approaches fromepidemiology and ecology to understand malaria as a zoonosis. Parasitology.2016;143:389–400.43. Rustam, Yasuda M, Tsuyuki S. Comparison of mammalian communities in ahuman-disturbed tropical landscape in East Kalimantan, Indonesia. MammStudy. 2012;37(4):299–311.44. Marchal V, Hill C. Primate crop-raiding: a study of local perceptions in fourvillages in North Sumatra, Indonesia. Primate Conserv. 2009;24:107–16.45. Linkie M, Dinata Y, Nofrianto A, Leader-Williams N. Patterns and perceptionsof wildlife crop raiding in and around Kerinci Seblat National Park, Sumatra.Anim Conserv. 2007;10(1):127–35.46. Brodie JF, Giordano A, Zipkin E, Bernard H, Mohd-Azlan J, Ambu L.Correlation and persistence of hunting and logging impacts on tropicalrainforest mammals. Conserv Biol. 2014;29(1):1–12.47. Son VD, Malaivijitnond S, Gotoh S, Tri CQ, Hung NV, Hoang LV, Trang TC,Hamada Y. Report on present distribution and status of macaques inVietnam. Lab Primate Newsl. 2009;48(1):10–4.48. Malaivijitnond S, Hamada Y. Current situation and status of long-tailedmacaques (Macaca fascicularis) in Thailand. Nat Hist J Chulalongkorn Univ.2008;8:185–204.49. Brotcorne F, Huynen MC, Savini T. Preliminary results on the behaviouralecology of a long-tailed macaque (Macaca fascicularis) population indisturbed urban habitats, Bangkok (Thailand). Folia Primatol. 2008;79(5):315.50. Hambali K, Ismail A, Zulkifli S, Munir Md-Zain B, Amir A. Human-macaqueconflict and pest behaviors of long-tailed macaques (Macaca fascicularis) inKuala Selangor Nature Park. Trop Nat Hist. 2012;12(2):189–205.Moyes et al. Parasites & Vectors  (2016) 9:242 Page 11 of 1251. Sha JC, Hanya G. Diet, activity, habitat use, and ranging of two neighboringgroups of food-enhanced long-tailed macaques (Macaca fascicularis). Am JPrimatol. 2013;75(6):581–92.52. Hamada Y, Kurita H, Goto S, Morimitsu Y, Malaivijitnond S, Pathonton F,Pathontone B, Kingsada P, Vongsombath C, Samouth F, et al. Distributionand present status of Macaques in Lao PDR. In: Nadler T, Rawson BM, ThinhVN, editors. Conservation primates in Indochina. Hanoi: Frankfurt ZoologicalSociety and Conservation International; 2010. p. 288.53. Coudrat C, Nekaris KA-I. Modelling niche differentiation of co-existing,elusive and morphologically similar species: A case study of four macaquespecies in Nakai-Nam Theun National protected area, Laos. Animals. 2013;3(1):45–62.54. Albert A, Hambuckers A, Culot L, Savini T, Huynen M-C. Frugivory and seeddispersal by northern pigtailed macaques (Macaca leonina), in Thailand. Int JPrimatol. 2013;34(1):170–93.55. San AM, Hamada Y. Distribution and current status of long-tailed macaques(Macaca fascicularis aurea) in Myanmar. In: Gumert MD, Fuentes A, Jones-Engel L, editors. Monkeys on the edge: ecology and management of long-tailed macaques and their interface with humans, vol. 60. Cambridge:Cambridge University Press; 2011. p. 45–71.56. Nijman V. The conservation status of mammals and birds in the ImawbumMountains, Kachin State, Northern Myanmar. Cambridge: Fauna and FloraInternational; 2015. 76.57. Riley EP, Priston NEC. Macaques in farms and folklore: Exploring the human-nonhuman primate interface in Sulawesi, Indonesia. Am J Primatol. 2010;72(10):848–54.58. Jones-Engel L, Engel GA, Schillact MA, Froehlich J, Paputungan U, Kyes RC.Prevalence of enteric parasites in pet macaques in Sulawesi, Indonesia. AmJ Primatol. 2004;62(2):71–82.59. Jones-Engel L, Engel GA, Schillaci MA, Pacheco M, Escalante A. Malarialmonkeys: reservoir for zoonotic infection? Am J Primatol. 2007;69(Suppl):40–1.60. Tan CH, Vythilingam I, Matusop A, Chan ST, Singh B. Bionomics of Anopheleslatens in Kapit, Sarawak, Malaysian Borneo in relation to the transmission ofzoonotic simian malaria parasite Plasmodium knowlesi. Malar J. 2008;7:52.61. William T, Rahman HA, Jelip J, Ibrahim MY, Menon J, Grigg MJ, Yeo TW,Anstey NM, Barber BE. Increasing incidence of Plasmodium knowlesi malariafollowing control of P. falciparum and P. vivax malaria in Sabah, Malaysia.PLoS Negl Trop Dis. 2013;7(1):e2026.62. Fornace KM, Abidin TR, Alexander N, Brock P, Grigg MK, Murphy A, William T,Menon J, Drakeley CJ, Cox J. Association between landscape factors and spatialpatterns of Plasmodium knowlesi infections in Sabah, Malaysia. Emerg InfectDis. 2016;22(2):201–8.63. Obsomer V, Defourny P, Coosemans M. Predicted distribution of majormalaria vectors belonging to the Anopheles dirus complex in Asia: Ecologicalniche and environmental influences. PLoS One. 2012;7(11):e50475.64. Maeno Y, Nguyen Tuyen Q, Culleton R, Kawai S, Masuda G, Nakazawa S,Marchand RP. Humans frequently exposed to a range of non-humanprimate malaria parasite species through the bites of Anopheles dirusmosquitoes in South-central Vietnam. Parasites Vectors. 2015;8:376.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:Moyes et al. Parasites & Vectors  (2016) 9:242 Page 12 of 12

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.52383.1-0366939/manifest

Comment

Related Items