Introduction
Climate change is happening more quickly and strongly than predicted, and the anthropic influence in this process is now clear (IPCC, Reference Pachauri and Meyer2014). Projections from several greenhouse gas emission scenarios agree on an increase of the mean earth surface temperature by the end of the 21st century, with continents heating more than oceans and high latitude regions heating more than the tropics. Longer and more frequent heat waves will probably occur, as well as more intense precipitation events in several regions (IPCC, Reference Pachauri and Meyer2014). Increased floods, droughts, fires, heat waves and air pollutants will directly impact human health. Indirect impacts on human health will arise from ecological disturbances and social responses to disruptions to agriculture, and to water and food supplies. Vector-borne diseases will also increase, compounded by human migrations towards endemic areas (Woodward et al., Reference Woodward, Smith, Campbell-Lendrum, Chadee, Honda, Liu, Olwoch, Revich, Sauerborn, Chafe, Confalonieri and Haines2014).
Vector-borne diseases are exceptionally sensitive to climate change because they emerge from complex transmission cycles involving several species of pathogens, vectors and hosts (Parham et al., Reference Parham, Waldock, Christophides, Hemming, Agusto, Evans, Feffermann, Gaff, Gumel, LaDeau, Lenhart, Mickens, Naumova, Ostfeld, Ready, Thomas, Velasco-Hernandez and Michael2015). Most disease vectors are arthropods, including insects and ticks. Climate change should, therefore, cause changes in disease distribution, density, seasonality and prevalence, and might prompt adaptation of vectors and hosts to new transmission cycles (Kovats et al., Reference Kovats, Campbell-Lendrum, Mcmichael, Woodward and Cox2001; Brooks & Hoberg, Reference Brooks and Hoberg2007; Rosenthal, Reference Rosenthal2009; Mills et al., Reference Mills, Gage and Khan2010).
The ecology of arthropod vectors should be impacted by climate change at three levels of biological organization: (i) at the individual level – being ectothermic organisms, vectors’ metabolism varies with daily fluctuations in temperature, which may affect physiological traits related to vector competence (Paaijmans et al., Reference Paaijmans, Heinig, Seliga, Blanford, Blanford, Murdock and Thomas2013) such as muscle activity (Harrison & Roberts, Reference Harrison and Roberts2000) and biting rates, although this latter influence is not entirely clear (Rogers & Randolph, Reference Rogers and Randolph2006; Ready, Reference Ready2013); (ii) at the population level – changes in climate should influence abundance, density, seasonality, survival rates, generation time, fecundity and dispersion ability, allowing vectors to colonize new habitats more efficiently (Mills et al., Reference Mills, Gage and Khan2010; Stange & Ayres, Reference Stange and Ayres2010; Eisen et al., Reference Eisen, Monaghan, Lozano-Fuentes, Steinhoff, Hayden and Bieringer2014); (iii) at the community level – parasite–vector interactions can be influenced by temperature (Hlavacova et al., Reference Hlavacova, Votypka and Volf2013), and new species of vectors or hosts can adapt to existing transmission cycles (Kovats et al., Reference Kovats, Campbell-Lendrum, Mcmichael, Woodward and Cox2001; Rosenthal, Reference Rosenthal2009; Parham et al., Reference Parham, Waldock, Christophides, Hemming, Agusto, Evans, Feffermann, Gaff, Gumel, LaDeau, Lenhart, Mickens, Naumova, Ostfeld, Ready, Thomas, Velasco-Hernandez and Michael2015).
Knowledge of vectors’ spatial distributions is essential to assess transmission risks in different regions. Predicting vector occurrence in specific regions is a challenge that many disease control programs must meet in order to plan and execute control interventions and adaptation measures more efficiently. With the popularization of GIS (geographic information systems), increasing availability of species occurrence data, disease information and environmental variables, various methods of spatial analysis and mathematical modelling have become common in the scientific literature. The methods that correlate these available data in order to predict species’ distributions are known as ecological niche models (ENMs) or species distribution models and have been widely used in studies of ecology, biogeography and conservation (Guisan & Zimmermann, Reference Guisan and Zimmermann2000; Guisan & Thuiller, Reference Guisan and Thuiller2005; Elith & Leathwick, Reference Elith and Leathwick2009). Recently, an increasing number of scientific articles have applied these models to study distributions of many medically important insect and tick species.
Ecological niche models are perhaps the most used methods to link climatic and environmental conditions to the distribution of species. In an ENM, an algorithm takes as input occurrence records of the studied species and calculates their relation with environmental variables, producing a surface of environmental suitability or probability of occurrence (Guisan & Zimmermann, Reference Guisan and Zimmermann2000; Franklin, Reference Franklin2010; Peterson et al., Reference Peterson, Soberón, Pearson, Anderson, Martínez-Meyer, Nakamura and Araújo2011). There are two basic approaches to apply an ENM in studies of vector-borne diseases. The first considers the entire transmission cycle and their ecological relationships as a ‘black box’, and analyses the geographical distribution of the disease occurrence, as if it were a single species (e.g. Nieto et al., Reference Nieto, Malone and Bavia2006; Yé et al., Reference Yé, Louis, Simboro and Sauerborn2007; Williams et al., Reference Williams, Fasina and Peterson2008; Arboleda et al., Reference Arboleda, Jaramillo-O and Peterson2009). This approach indirectly groups all component species of the transmission cycle, as well as their environmental needs and ecological interactions, losing, therefore, important details of the transmission process. However, the occurrence of the disease is often the only information available, and this becomes the only modelling option. The second approach is to model each species from the transmission cycle individually, and evaluate areas of co-occurrence afterwards. This approach offers the opportunity to distinguish different reasons for the presence or absence of disease transmission in certain locations. For example, the disease may be absent due to the lack of the pathogen, an appropriate vector or a reservoir host (Peterson et al., Reference Peterson, Soberón, Pearson, Anderson, Martínez-Meyer, Nakamura and Araújo2011). Areas with the presence of only vectors and competent hosts may be treated as vulnerable – a particularly important situation nowadays, when species are artificially transported by humans and new diseases emerge in areas where they would not naturally occur (Komar, Reference Komar2003; Ready, Reference Ready2008, Reference Ready2010; Daszak et al., Reference Daszak, Zambrana-Torrelio, Bogich, Fernandez, Epstein, Murray and Hamilton2013).
Comparative studies show that most of uncertainty in ENM comes from using different modelling algorithms (Buisson et al., Reference Buisson, Thuiller, Casajus, Lek and Grenouillette2009; Diniz-Filho et al., Reference Diniz-Filho, Bini, Rangel, Loyola, Hof, Nogués-Bravo and Araújo2009; Elith & Graham, Reference Elith and Graham2009). With the wide variety of methods, it is an additional challenge to interpret and compare the results of studies on vector distributions, so that they can be effectively used in control programs. Here we review the future projections of disease vectors produced by ENMs, and assess trends and limitations of the methods applied.
Methods
We performed a systematic review of the literature using four online databases: (i) Web of Science (http://isiwebofknowledge.com); (ii) Scopus (http://www.hub.sciverse.com); (iii) Pubmed (http://www.ncbi.nlm.nih.gov/pubmed); and (iv) Scientific Electronic Library Online (SciELO) (http://www.scielo.org). The Web of Science is the most comprehensive database of peer-reviewed articles published in English, as well as being the most used in systematic reviews (Falagas et al., Reference Falagas, Pitsouni, Malietzis and Pappas2008; Gavel & Iselid, Reference Gavel and Iselid2008). However, Scopus covers a larger number of journals that publish articles in languages other than English (Falagas et al., Reference Falagas, Pitsouni, Malietzis and Pappas2008; Gavel & Iselid, Reference Gavel and Iselid2008). PubMed is the most frequently consulted source for information in the biomedical field (Falagas et al., Reference Falagas, Pitsouni, Malietzis and Pappas2008). The SciELO database, although less comprehensive, includes many Latin American journals that are not included in the other consulted databases.
Searches were conducted in March 2015, through different combinations of the following key words: ‘ecologic* niche model*’ ‘species distribution model*’, ‘climat* model*’, ‘vector’, ‘disease’. The initial results (N = 572) were limited to articles published until 2014 that applied ENMs to predict areas of occurrence or environmental suitability of arthropods vectors of diseases. Articles that used models to explain the relationship of the vectors with environmental variables, without predictive mapping, were excluded from the analysis. Studies with models based only on the occurrence of disease or risk maps generated without vector information were also discarded. After removing duplicates and refining selections, 146 articles were reviewed (Table S1).
The articles were described under the following categories: vector species and main associated disease; study area; types of biological data; types of environmental data; applied method; inclusion of future projections (Table S1). Studies including future projections were analysed in greater detail in relation to biological data (number of records, data source), environmental data (number of variables, approximate spatial resolution), methods (algorithm employed, use of ensemble models based on different algorithms) and future projections (years, general circulation model, climate change scenario) (Table S2). The main results of the future projections were summarized by vector group and further described.
Results and discussion
Application of different modelling methods
Seventeen different modelling methods were used to predict vector distributions. The most common was the maximum entropy algorithm (MaxEnt, 43 articles) (Phillips et al., Reference Phillips, Anderson and Schapire2006), followed by generalized linear models (GLM, 34) (Guisan et al., Reference Guisan, Edwards and Hastie2002), the genetic algorithm for rule set prediction (GARP, 25) (Stockwell, Reference Stockwell1999), and discriminant analysis (12) (Rogers et al., Reference Rogers, Hay and Packer1996) (fig. 1, Table S1). Other methods were less frequently applied, such as CLIMEX (5) (Sutherst & Maywald, Reference Sutherst and Maywald1985), ENFA (3) (Hirzel et al., Reference Hirzel, Hausser, Chessel and Perrin2002), BRT (2) (Elith et al., Reference Elith, Leathwick and Hastie2008), BIOCLIM (1) (Booth et al., Reference Booth, Nix, Busby and Hutchinson2014) and Random Forests (1) (Breiman, Reference Breiman2001) (fig. 1, Table S1). For comprehensive descriptions of ENM algorithms, see Franklin (Reference Franklin2010) and Peterson et al. (Reference Peterson, Soberón, Pearson, Anderson, Martínez-Meyer, Nakamura and Araújo2011).
Methods applied in the literature of ecological niche modelling of arthropod vectors of diseases.
The predictive performance of MaxEnt has exceeded other algorithms in several comparative studies (Elith et al., Reference Elith, Graham, Anderson, Dudík, Ferrier, Guisan, Hijmans, Huettmann, Leathwick, Lehmann, Li, Lohmann, Loislelle, Manion, Moritz, Nakamura, Nakazawa, Overton, Townsend Peterson, Phillips, Richardson, Scachetti-Pereira, Schapire, Soberón, Williams, Wisz and Zimmermann2006; Foley et al., Reference Foley, Klein, Kim, Sames, Wilkerson and Rueda2009, Reference Foley, Klein, Kim, Brown, Wilkerson and Rueda2010; Larson et al., Reference Larson, Degroote, Bartholomay and Sugumaran2010; Arboleda et al., Reference Arboleda, Jaramillo-O and Peterson2012). In addition, its popularity can probably be explained by the fact that it is implemented in free software with a user-friendly interface, good documentation and many options for parameterization. Generalized linear models were the second-most frequent method because they offer more flexibility than machine learning algorithms (e.g. MaxEnt and GARP), thus improving model fit and ecological interpretations of parameters (Franklin, Reference Franklin2010). Also noteworthy is the use of CLIMEX, a mechanistic (process-based) algorithm. Mechanistic models are based on vector's biological processes, such as duration of life cycle, biting rates, dispersal ability, temperature limits for larvae development, etc. The inclusion of this type of data improves the biological meaning of models, but they require solid empirical knowledge about the vectors’ physiology, which makes parameterization a challenge (Kearney & Porter, Reference Kearney and Porter2009; Dormann et al., Reference Dormann, Schymanski, Cabral, Chuine, Graham, Hartig, Kearney, Morin, Römermann, Schröder and Singer2012; Fischer et al., Reference Fischer, Thomas, Neteler, Tjaden and Beierkuhnlein2014).
Models produced by different algorithms may have dissimilar, even contrasting outputs (Dormann et al., Reference Dormann, Purschke, Márquez, Lautenbach and Schröder2008; Diniz et al., Reference Diniz-Filho, Bini, Rangel, Loyola, Hof, Nogués-Bravo and Araújo2009; Elith & Graham, Reference Elith and Graham2009; Li & Wang, Reference Li and Wang2013). Independent evaluations have often been unable to identify a single recommended algorithm for all circumstances (Elith et al., Reference Elith, Graham, Anderson, Dudík, Ferrier, Guisan, Hijmans, Huettmann, Leathwick, Lehmann, Li, Lohmann, Loislelle, Manion, Moritz, Nakamura, Nakazawa, Overton, Townsend Peterson, Phillips, Richardson, Scachetti-Pereira, Schapire, Soberón, Williams, Wisz and Zimmermann2006; Elith & Graham, Reference Elith and Graham2009; Li & Wang, Reference Li and Wang2013; Qiao et al., Reference Qiao, Soberón and Peterson2015). An alternative to avoid the choice of a particular method is to test models produced by a set of algorithms (Qiao et al., Reference Qiao, Soberón and Peterson2015) and combine their results as an ensemble model (Araújo & New, Reference Araújo and New2007; Marmion et al., Reference Marmion, Parviainen, Luoto, Heikkinen and Thuiller2009). With a set of models produced by a number of algorithms, uncertainty can be properly quantified, thus improving the study's result (Pearson et al., Reference Pearson, Thuiller, Araújo, Martinez-Meyer, Brotons, Mcclean, Miles, Segurado, Dawson and Lees2006; Owens et al., Reference Owens, Campbell, Dornak, Saupe, Barve, Soberón, Ingenloff, Lira-Noriega, Hensz, Myers and Towsend Peterson2013; Qiao et al., Reference Qiao, Soberón and Peterson2015). The use of multiple algorithms was present in over 70% of general ENM studies published recently (Guillera-Arroita et al., Reference Guillera-Arroita, Lahoz-Monfort, Elith, Gordon, Kujala, Lentini, McCarthy, Tingles and Wintle2015), but it was under-represented in ENM of disease vectors for the same period (approximately 10%). This represents a significant delay in disease vector studies in relation to what is currently being published.
A good example of the multiple algorithm approach was a comparison between models produced by BIOCLIM, DOMAIN (Carpenter et al., Reference Carpenter, Gillison and Winter1993), GARP, GLM (logistic regression) and MaxEnt to identify areas of high density of Aedes mosquitoes in Bermuda (Khatchikian et al., Reference Khatchikian, Sangermano, Kendell and Livdahl2011). The results varied between the different algorithms, but since GLM and MaxEnt performed better, both were used to predict risk areas of mosquito infestations (Khatchikian et al., Reference Khatchikian, Sangermano, Kendell and Livdahl2011). Another example was a study of the distribution patterns of natural breeding sites of A. aegypti in Colombia, where models produced by GARP had fewer omission errors than those produced by MaxEnt (Arboleda et al., Reference Arboleda, Jaramillo-O and Peterson2012). Models produced by MaxEnt performed better in certain regions, although areas predicted as suitable by the two algorithms coincided closely. The two algorithms were combined into an ensemble model, where coincident areas were considered suitable with greater confidence. The combination of methods improved the detection of natural breeding sites, allowing the optimization of effort and financial investment in dengue control programs in the region (Arboleda et al., Reference Arboleda, Jaramillo-O and Peterson2012).
Future projections of vector distributions
Over 700 vector species were studied in the 146 reviewed papers, including mostly mosquitoes (63 articles) and sand flies (29), followed by works on kissing bugs (18), biting midges (17), ticks (14), tsetse flies (3), fleas (1) and water bugs (1) (Table S1). The geographic extent of the reviewed studies varied from local to global (Table S1). The 31 studies with future ENM projections mostly point to expansions in response to climate change scenarios, accompanied by poleward shifts (table 1). This trend is being observed for several taxonomic groups, where long-term field studies demonstrate recent species’ movements towards higher latitudes and higher altitudes in response to climate change (Hickling et al., Reference Hickling, Roy, Hill, Fox and Thomas2006; Stange & Ayres, Reference Stange and Ayres2010; Chen et al., Reference Chen, Hill, Ohlemüller, Roy and Thomas2011). There is, however, a noteworthy methodological issue in about half of the reviewed studies (table 1). When projecting into future scenarios, models should be trained with the full-known distribution of the species. If only a subset of the realized niche is used, future predictions may underestimate environmentally suitable areas and quantifications of range changes become questionable (Pearson & Dawson, Reference Pearson and Dawson2003; Guisan & Thuiller, Reference Guisan and Thuiller2005; Araújo & Peterson, Reference Araújo and Peterson2012). An additional source of uncertainty in future forecasts is the extrapolation of models into climatic conditions that do not presently exist (Fitzpatrick & Hargrove, Reference Fitzpatrick and Hargrove2009). Some ENM algorithms have standard ways of controlling extrapolation, such as MaxEnt, by limiting output values to the range of environmental variables under which the model was calibrated (Phillips et al., Reference Phillips, Anderson and Schapire2006). Alternatively, out-of-range values can be masked directly in model predictions (Owens et al., Reference Owens, Campbell, Dornak, Saupe, Barve, Soberón, Ingenloff, Lira-Noriega, Hensz, Myers and Towsend Peterson2013; Carvalho et al., Reference Carvalho, Rangel, Ready and Vale2015).
Overview of the future projections of the distributions of arthropod vectors of diseases.
Aedes aegypti and Aedes albopictus (Diptera: Culicidae)
The main vector of dengue, A. aegypti, is currently distributed throughout most tropical regions of the world. Projections of its global distribution showed that most areas that are currently occupied should remain climatically favourable for its occurrence in 2030 and 2070, while new areas will become suitable for its range expansion, such as the Australian outback, the Arabian Peninsula, southern Iran and parts of North America (Khormi & Kumar, Reference Khormi and Kumar2014). Further projections for the near future indicate that suitable macroclimatic conditions for this vector should begin to expand between 2010 and 2039 (Capinha et al., Reference Capinha, Rocha and Sousa2014). In Brazil, models predict a contraction of its range in the northern and northeastern regions, accompanied by a probable expansion in the south by 2050 (Cardoso-Leite et al., Reference Cardoso-Leite, Vilarinho, Novaes, Tonetto, Vilardi and Guillermo-Ferreira2014). Since A. aegypti is a vector with high adaptability to urban environments, its local distribution is also influenced by the occurrence of artificial breeding sites such as water tanks and swimming pools. In Australia, models based only on climatic variables failed to detect locations of its known occurrence and of human cases of dengue (Beebe et al., Reference Beebe, Cooper, Mottram and Sweeney2009). This inconsistency was attributed to human behaviour, as residents began to store water during a regional drought attributed to climate change (Beebe et al., Reference Beebe, Cooper, Mottram and Sweeney2009). The study pointed out, therefore, not only the local-scale limitations of ENM, but also the importance of implementing climate change adaptation measures that are compatible with disease control programs (Beebe et al., Reference Beebe, Cooper, Mottram and Sweeney2009). Despite not using the full-known distribution of the vector in model training, predictions of future range contraction in Brazil (Cardoso-Leite et al., Reference Cardoso-Leite, Vilarinho, Novaes, Tonetto, Vilardi and Guillermo-Ferreira2014) and of future inland expansion in Australia (Beebe et al., Reference Beebe, Cooper, Mottram and Sweeney2009) were similar to those predicted by models based on its global distribution (Capinha et al., Reference Capinha, Rocha and Sousa2014; Khormi & Kumar, Reference Khormi and Kumar2014).
In contrast to the highly anthropophilic A. aegypti, the Asian tiger mosquito A. albopictus prefers less disturbed environments and has predominantly zoophilic habits, participating in sylvatic transmission cycles of a number of arboviruses, such as chikungunya, yellow fever and dengue. Despite having relatively lower importance in human disease transmission than A. aegypti, the distribution of A. albopictus has been much studied because it is considered the most invasive mosquito species in the world (Benedict et al., Reference Benedict, Levine, Hawley and Lounibos2007; Medley, Reference Medley2010; Porretta et al., Reference Porretta, Mastrantonio, Bellini, Somboon and Urbanelli2012). Its original distribution in Southeast Asia has expanded in recent decades to various countries in the Americas, Africa and Europe, mostly through cargo transportation (Reiter & Sprenger, Reference Reiter and Sprenger1987; Tatem et al., Reference Tatem, Hay and Rogers2006). In Europe, the species is currently established in the Mediterranean region, where local vector populations are already expanding (Roiz et al., Reference Roiz, Neteler, Castellani, Arnoldi and Rizzoli2011). Future projections of ENM point to an increase of climate suitability areas for A. albopictus in central and western parts of Europe by 2040, with eastern areas becoming suitable from 2070 onwards (Fischer et al., Reference Fischer, Thomas, Niemitz, Reineking and Beierkuhnlein2011c ). Based on these projections, an assessment of the main cargo shipment routes concluded that certain areas of the continent, such as Rotterdam, Hamburg and Antwerp, have the dangerous combination of high incoming cargo from countries where A. albopictus occurs and high future climatic suitability for the vector (Thomas et al., Reference Thomas, Tjaden, Van Den Bos and Beierkuhnlein2014). In Australia, where there are currently no records of A. albopictus, niche models based on the global distribution of the vector show that the coastal region is climatically suitable for its establishment, with projections for the coming decades indicating expansion of this suitable area towards the interior of the country (Hill et al., Reference Hill, Axford and Hoffmann2014).
Anopheles spp. (Diptera: Culicidae)
The distributions of two malaria vectors in Sub-Saharan Africa, Anopheles gambiae and An. arabiensis, will also likely expand southwards and southeastwards, according to ENMs involving climate change scenarios (Peterson, Reference Peterson2009; Fuller et al., Reference Fuller, Parenti, Hassan and Beier2012; Tonnang et al., Reference Tonnang, Kangalawe and Yanda2010, Reference Tonnang, Tchouassi, Juarez, Igweta and Djouaka2014). By adding mosquito survival rates to niche models, it was concluded that East African countries will have greater climatic suitability for these vectors in the coming decades than West African countries (Tonnang et al., Reference Tonnang, Tchouassi, Juarez, Igweta and Djouaka2014). Although the models predict local regions of both increase and decrease of climatic suitability for the vectors, 11–30% fewer people should be exposed to the vectors in the coming decades, as seen by overlaying model predictions and human distribution (Peterson, Reference Peterson2009). A more recent study pointed to a contraction of over half of the distribution area of An. arabiensis in West African countries (Drake & Beier, Reference Drake and Beier2014). The overall contraction of the full range of the vector might erroneously suggest less exposure to vector-borne diseases with climate change. The association with human distribution demonstrates the caution needed when interpreting predictions of ENMs of vectors. Vector occurrence per se does not necessarily implies higher risk of disease transmission, and a closer look at other risk factors is needed.
Lutzomyia spp. and Phlebotomus spp. (Diptera: Psychodidae)
Leishmaniases are neglected tropical diseases widely distributed in 98 countries, with approximately 0.2–0.4 million cases of visceral leishmaniasis and 0.7–1.2 million cases of cutaneous leishmaniasis occurring every year (Alvar et al., Reference Alvar, Vélez, Bern, Herrero, Desjeux, Cano, Jannin and den Boer2012). The vectors of leishmaniases, sand flies, are classified in two genera according to their distributions: Lutzomyia in the Americas and Phlebotomus in other continents. Areas climatically suitable for the South American vectors L. whitmani, L. intermedia and L. migonei should expand by the year 2050 (Peterson & Shaw, Reference Peterson and Shaw2003). Expansion areas are located in different regions of the continent, but their most evident direction is south, where L. whitmani will have larger suitability areas than the other two vectors (Peterson & Shaw, Reference Peterson and Shaw2003). In contrast, in Colombia, future projections from regional distribution models indicate reduction of the total predicted area of occurrence of L. longipalpis and L. evansi associated with changes in their altitudinal distribution (González et al., Reference González, Paz and Ferro2014). Unfortunately, failure to consider the full distribution of L. longipalpis might have produced biased predictions for Colombia (González et al., Reference González, Paz and Ferro2014), because the vector occupies a broad range of latitudes from Mexico to Argentina (World Health Organization, 2010).
The vectors L. anthophora and L. diabolica, currently distributed in Mexico and the USA, are projected to expand northwards (González et al., Reference González, Wang, Strutz, González-Salazar, Sánchez-Cordero and Sarkar2010). These projections were associated with predictions of the distributions of rodent hosts and human populations, and indicated that the expected number of people exposed to leishmaniases in North America will at least double by 2080 (González et al., Reference González, Wang, Strutz, González-Salazar, Sánchez-Cordero and Sarkar2010). Future northward expansions of suitable areas to leishmaniasis vectors are also expected for 27 of 28 species with current occurrence in Mexico, Guatemala, Belize, USA and Canada, the exception being L. vexator (Moo-Llanes et al., Reference Moo-Llanes, Ibarra-Cerdeña, Rebollar-Téllez, Ibáñez-Bernal, González and Ramsey2013). However, the predictions for species whose distributions include South America, such as L. longipalpis and L. shannoni, should be interpreted with caution because the models were only calibrated with data from Canada, USA, Mexico, Guatemala and Belize (Moo-Llanes et al., Reference Moo-Llanes, Ibarra-Cerdeña, Rebollar-Téllez, Ibáñez-Bernal, González and Ramsey2013).
Europe is currently on alert for the emergence of leishmaniasis and the expansion of its vectors, especially in the countries in central regions of the continent, predicted to become increasingly climatically similar to the Mediterranean region, where there are endemic areas of these diseases (Ready, Reference Ready2008; Medlock et al., Reference Medlock, Hansford, Van Bortel, Zeller and Alten2014). In a region of canine leishmaniasis in Spain, an increase in the abundance of P. ariasi in higher altitude areas was observed, pointing out to a possible migration of the vector to these areas in response to rising temperatures (Gálvez et al., Reference Gálvez, Descalzo, Miró, Jiménez, Martín, Sandos-Brandao, Guerrero, Cubero and Molina2010). Future projections predict expansions of the range and increase of local densities of both P. ariasi and P. perniciosus in the 21st century (Gálvez et al., Reference Gálvez, Descalzo, Guerrero, Miró and Molina2011); however, models were restricted to Spain, which represents only part of the range of both species at the Mediterranean region (World Health Organization, 2010). In Germany, Austria and Switzerland, there are predicted areas of increased climate suitability for five species of Phlebotomus, but most are unlikely to be reached by the vectors by the end of this century due to their limited dispersal ability (Fischer et al., Reference Fischer, Moeller, Thomas, Naucke and Beierkuhnlein2011a ). This finding was reinforced by later field sampling in the region of Bavaria, southern Germany, where no sand flies were caught (Haeberlein et al., Reference Haeberlein, Fischer, Thomas, Schleicher, Beierkuhnlein and Bogdan2013). However, field studies show that several species of Phlebotomus from the Mediterranean region already have records of the expansion of their distributions towards central Europe (Maroli et al., Reference Maroli, Rossi, Baldelli, Capelli, Ferroglio, Genchi, Gramicia, Mortarino, Pietrobelli and Gradoni2008; Medlock et al., Reference Medlock, Hansford, Van Bortel, Zeller and Alten2014).
Culicoides spp. (Diptera: Ceratopogonidae)
Bluetongue disease, a zoonotic infection transmitted by Culicoides spp. (biting midges) to various ruminants, has important economic impacts in temperate zones of Europe, Africa and the Americas. Some authors suggest that in the Mediterranean region there is evidence of northward expansion of C. imicola in recent decades (Purse et al., Reference Purse, Mellor, Rogers, Samuel, Mertens and Baylis2005), while others refute this (Conte et al., Reference Conte, Gilbert and Goffredo2009). Future expansions of C. imicola in climate change scenarios are predicted for most of its occurrence areas in the northern hemisphere (mainly central and western Europe and the USA) and some contraction areas in Africa (Guichard et al., Reference Guichard, Guis, Tran, Garros, Balenghien and Kriticos2014). In Europe, their distribution is currently known in the Iberian Peninsula, with future climatically suitable areas predicted in the northwest direction, in climate change scenarios (Wittmann et al., Reference Wittmann, Mellor and Baylis2001). In Spain, niche models of wild hosts of Bluetongue virus (deer and wild boar) were used as predictors of the occurrence of C. imicola, in addition to other environmental variables, showing that in the near future (2011–2040), its predicted distribution will not suffer many changes, but its abundance is expected to increase in currently occupied areas (Acevedo et al., Reference Acevedo, Ruiz-Fons, Estrada, Márquez, Miranda and Gortázar2010).
Triatoma spp. (Hemiptera: Reduviidae)
Chagas disease, also known as American trypanosomiasis, is transmitted by many species of kissing bugs from Triatominae subfamily. It was originally restricted to Latin America, but in past decades it has been detected in the USA, Canada, European and Asian countries, due mostly to human migration from endemic areas Schmunis & Yadon, Reference Schmunis and Yadon2010). In Brazil, Triatoma brasiliensis, a species complex (Monteiro et al., Reference Monteiro, Donnelly, Beard and Costa2004), is considered the main vector in the northeast region (Monteiro et al., Reference Monteiro, Donnelly, Beard and Costa2004). Future projections of its distribution indicate few areas of both expansion and contraction, so its distribution may remain stable, at least until 2050 (Costa et al., Reference Costa, Dornak, Almeida and Peterson2014). In contrast, ENMs of two vectors of Chagas disease in the USA, T. gerstaeckeri and T. sanguisuga, predict northwards expansions of their distributions in response to climate change in 2050 (Garza et al., Reference Garza, Arroyo, Casillas, Sanchez-Cordero, Rivaldi and Sarkar2014).
Ixodes spp. (Acari: Ixodida)
Several species of Ixodes ticks are involved in the transmission of Lyme disease, which is the most prevalent vector-borne disease in the USA and Europe. It is vectored by I. scapularis and I. pacificus in North and Central America, and by I. persulcatus and I. ricinus in Europe and Asia (Lane et al., Reference Lane, Piesman and Burgdorfer1991).
The distribution of I. ricinus in Europe may nearly double by 2080 (Porretta et al., Reference Porretta, Mastrantonio, Amendolia, Gaiarsa, Epis, Genchi, Bandi, Otranto and Urbanelli2013). This predicted expansion includes areas north and east of its current range, reaching the northernmost regions of Eurasia, such as Sweden and Russia (Porretta et al., Reference Porretta, Mastrantonio, Amendolia, Gaiarsa, Epis, Genchi, Bandi, Otranto and Urbanelli2013). A model developed from a subset of its distribution records showed overall similar future predictions for Europe, with some local differences in the Iberian Peninsula and Scandinavia (Boeckmann & Joyner, Reference Boeckmann and Joyner2014). In the USA, models indicate current greater probability of occurrence of I. scapularis in the Gulf of Mexico, and future projections point to relative stability in its range by 2050 (Feria-Arroyo et al., Reference Feria-Arroyo, Castro-Arellano, Gordillo-Perez, Cavazos, Vargas-Sandovál, Grover, Torres, Medina, Pérez de León and Esteve-Gassent2014).
Further considerations on niche models of disease vectors
Vector occurrence data often present spatial bias towards endemic areas where disease surveillance programs are active. In addition, having presence and absence data that are required for some ENM algorithms is rarely the case when studying disease vectors. Most studies that used absence data based on field studies were restricted to regional and local scales, due to the inherent limitations of sampling effort (Eisen et al., Reference Eisen, Eisen and Lane2006; Mushinzimana et al., Reference Mushinzimana, Munga, Minakawa, LI, Feng, Bian, Kitron, Schmidt, Beck, Zhou, Githeko and Yan2006; Reiter & Lapointe, Reference Reiter and Lapointe2007; Khatchikian et al., Reference Khatchikian, Sangermano, Kendell and Livdahl2011; Cardo et al., Reference Cardo, Vezzani, Rubio and Carbajo2014). Absence data can be replaced by pseudo-absences generated according to several criteria (Lobo & Tognelli, Reference Lobo and Tognelli2011; Senay et al., Reference Senay, Worner and Ikeda2013). Real absence data, however, can also be a source of bias in model outputs if they are not treated appropriately. After all, a species may be absent from a sampled region for various reasons besides the lack of environmental suitability, such as dispersion barriers, historical factors or biotic interactions (Lobo et al., Reference Lobo, Jiménez-Valverde and Hortal2010). In a modelling exercise to test different absence datasets of C. imicola, the removal of false absences improved all model outputs (Peters et al., Reference Peters, De Baets, Van Doninck, Calvete, Lucientes, De Clercq, Ducheyne and Verhoest2011).
Most ENMs are correlative approaches based on abiotic factors; they do not consider species’ dispersion (Guisan & Zimmermann, Reference Guisan and Zimmermann2000; Barve et al., Reference Barve, Barve, Jiménez-Valverde, Lira-Noriega, Maher, Peterson, Soberón and Villalobos2011). Thus, knowledge of vector ecology becomes essential for interpretation of model outputs. Accessible localities in climatic suitability areas can be either hypothesised a priori (Barve et al., Reference Barve, Barve, Jiménez-Valverde, Lira-Noriega, Maher, Peterson, Soberón and Villalobos2011; Carvalho et al., Reference Carvalho, Rangel, Ready and Vale2015) or mapped a posteriori for vectors with limited dispersal ability, such as sand flies (Fischer et al., Reference Fischer, Moeller, Thomas, Naucke and Beierkuhnlein2011a ). In contrast, ticks’ dispersion is facilitated by their hosts’ movements, favouring their range expansion in suitable areas (Porretta et al., Reference Porretta, Mastrantonio, Amendolia, Gaiarsa, Epis, Genchi, Bandi, Otranto and Urbanelli2013).
Health data are commonly grouped into administrative areas, such as municipalities, districts, states or countries. Automatically converting vector records from this format to point localities can generate positional errors, depending on the spatial resolution of the study, which might lead to wrong estimates of the species–environment relationship (Naimi et al., Reference Naimi, Hamm, Groen, Skidmore and Toxopeus2014). Even if the vector records for an ENM are aggregated into area units, they can be analysed using statistical methods, considering the spatial limitations of model predictions. This approach was applied in a GLM (logistic regression) of the environmental suitability of L. whitmani, cutaneous leishmaniasis vector in the state of Mato Grosso, Brazil, where both vector occurrence and environmental data were grouped at the municipal level (Zeilhofer et al., Reference Zeilhofer, Kummer, dos Santos, Ribeiro and Missawa2008).
To correctly interpret future ENM projections for disease vectors, it is important to remember that a vector's distribution represents only a fraction of the factors that determine human vector-borne diseases. Even if vectors, pathogens and hosts co-exist in a location, the disease might not become endemic for several reasons. Human social factors play an important role in disease establishment, such as migration, urbanization, population immunity and effectiveness of health systems (Gage et al., Reference Gage, Burkot, Eisen and Hayes2008; Barcellos et al., Reference Barcellos, Monteiro, Corvalán, Gurgel, Carvalho, Artaxo, Hacon and Ragoni2009). For example, the incidence of malaria has declined since 1900, mainly due to effective control (Gething et al., Reference Gething, Smith, Patil, Tatem, Snow and Hay2010). However, in the areas of predicted expansion of distribution of malaria vectors in Africa there is more poverty and fewer resources to control the disease, which are important determinants of transmission risk (Peterson, Reference Peterson2009). International travel has contributed to increased numbers of imported cases of dengue in the USA and Europe (Gardner et al., Reference Gardner, Fajardo, Waller, Wang and Sarkar2012). Chagas disease, a chronic and silent infection currently treated as an emergent vector-borne disease in southern USA, may have been established in the region for over 70 years (Garcia et al., Reference Garcia, Woc-Colburn, Aguilar, Hotez and Murray2015), so the predicted expansion of vectors may increase transmission risk (Garza et al., Reference Garza, Arroyo, Casillas, Sanchez-Cordero, Rivaldi and Sarkar2014). Canine leishmaniasis transmission cycles, known to precede human outbreaks of the disease have been recorded in areas with no records of human cases, not only in European countries (Ready, Reference Ready2010), but also in the USA and Canada (Duprey et al., Reference Duprey, Steurer, Rooney, Kirchhoff, Jackson, Rowton and Schantz2006). These and other evidence points to the need for a multidisciplinary view of the impacts of climate change on vector-borne diseases.
Conclusions
Changes in the geographical distribution of vectors are expected with climate change, therefore impacting the spatial epidemiology of vector-borne diseases. Tropical regions of the world are currently occupied by many vector species, however future projections indicate poleward increases of suitable climates for their occurrence. These are the scenarios for Mediterranean vectors of several arboviruses, leishmaniasis, bluetongue disease and tick-borne infections, which are expected to find climatically suitable areas in central Europe for their expansions by the end of this century. In Sub-Saharan Africa, malaria vectors are expected to shift their distributions southward and eastward, losing climatic suitability in western countries in the process. Leishmaniasis vectors from tropical America are projected to expand their ranges both northwards and southwards in temperate zones, while inland Australia should increase in climatic suitability for mosquitoes.
The results discussed here are for the distribution of vectors only, which are a fraction of the determinants of the occurrence of these diseases. These likely vector expansions will only translate into increased risk of human disease if they are accompanied by hosts and parasites themselves. Human social factors and control efforts also play important roles in transmission risk. It is recommended that entomological monitoring activities are made, especially in the areas projected to become suitable for the occurrence of these vectors. Long-term monitoring studies can contribute substantially to the knowledge of the ecology of these species and how their distributions change in response to climate change.
Adoption of multiple ENM methods to study disease vector distributions is slow relative to the general ENM literature. Another concern is the lack of consideration of the full-known current distribution of the target species on models that include future projections; about half of the reviewed studies had this issue, potentially leading to questionable predictions. An extra effort from authors is necessary in order to better understand the details of these methods so that models are produced with greater reliability and a clear description of their uncertainties. With this, these studies can support disease control policies more efficiently.
Supplementary Material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007485316001097
