Background
The interactions between a species and its environment are reflected in the distribution of its abundance in both space and time [
1]. Species are expected to be non-randomly distributed across different ecological settings, as a result of their specific ecological requirements and tolerance towards deviations from their optimal conditions [
2,
3]. Predictions of species geographic distributions can be based upon mathematical models relating field observations of occurrences to a set of environmental variables [
4,
5]. This kind of approach has been used to explore ecological niche requirements and to predict the potential distribution of a focal species [
6]. Such predictions can be used to tackle a wide range of issues such as conservation of biodiversity, the management of species of economic interest, or evaluation of the risks linked with biological invasions [
7‐
10]. Species distribution models are also gaining interest as a tool to evaluate and/or predict the risk of exposure to infectious diseases and their vectors, such as malaria [
11‐
14], Chagas disease [
15] or dengue [
16]. Risk maps have been produced by correlating geo-referenced epidemiological and environmental data to describe, explain and predict malaria risk at localities where epidemiological data are not available [
11,
17,
18]. Mosquito life-history traits, such as growth rates and survival and the duration of the sporogonic cycle of
Plasmodium in its vector, are strongly dependent upon temperature and moisture conditions on the ground. Thus, eco-climatic profiles inferred from remotely sensed images can be used as predictors of mosquito distribution patterns and average levels of transmission of malaria parasites by these vectors [
12].
Malaria transmission dynamic is highly variable throughout Africa. These variations mirror, at least to some extent, the great heterogeneity of eco-climatic settings present across sub-Saharan Africa [
19]. In this continent, about twenty out of 140 anopheline species have been incriminated in malaria transmission [
20,
21]. However, only five species are responsible for more than 95% of the overall transmission, and are therefore considered the major malaria vectors in Africa:
Anopheles gambiae,
Anopheles arabiensis,
Anopheles funestus,
Anopheles moucheti, and
Anopheles nili [
19,
21]. The remaining 5% is due to "secondary" malaria vectors of local importance. Differences in ecological requirements, longevity and feeding behaviour (e.g. anthropophily) account for the different roles played by major and secondary vectors in malaria transmission [
22]. Whereas variations in longevity and anthropophily within and between vectors species have been documented under a wide range of settings, qualitative and quantitative assessments of species' ecological requirements are still few, even for major vector species [
19,
23].
This paper focuses on the determination of ecological requirements for malaria vectors in Cameroon, a country in Central Africa covering a wide range of ecological and climatic domains. This great environmental heterogeneity increases the diversity of the malaria transmission system, with as much as 48 anophelines species reported [
24‐
26], among which 17 have been found infected with human malaria parasites [
22,
27‐
30]. Geographical Information Systems and Ecological Niche Factor Analysis (ENFA) [
3] were employed to build predictive habitat suitability maps for the five major malaria vectors on a country-wide scale, and to compare their respective ecological requirements and niche parameters. In addition, the ecological habitat profiles for the 10 most common anopheline species found in Cameroon were described using Canonical Correspondence Analysis (CCA). This study improves current knowledge of malaria vector distribution in Cameroon and highlights relevant similarities and differences in ecological requirements of the different mosquito species present in this area. The relevance of these findings for malaria vector control in Africa is discussed.
Discussion
One of the main aims of ecologists is to characterize the distribution and abundance of animal populations [
57]. Knowledge of where an organism lives is a fundamental requisite for the understanding of its ecology and more detailed analyses about its biology. Nowadays, the availability of accurate environmental data over large spatial extents, together with inexpensive and powerful ways to manipulate and analyse such data, has spurred the development of analytical techniques aimed at predicting species environmental requirements, from which their geographic distribution can be inferred [
58].
In tropical Africa, several studies have addressed, by different modelling approaches, the geographic distribution of malaria vectors with the objective to predict malaria transmission risk at the continental scale [
12,
23,
59]. Most studies, however, have relied on collation of published entomological data gathered for purposes different from that of studying the vectors' geographic distribution [
23,
59,
60]. The 'training set' of locations on which such species distribution models are based are generally a biased sample of the sites, where malaria vectors may be effectively present in nature. For example, the absence of a mosquito species in a given locality is likely to reflect the lack of medical entomologists that worked previously in that area, the nature of the collection methods, the seasonal pattern of species abundance, and other environmental or historical accidents, highlighting major limitations of distribution models relying on both accurate presence
and absence data. At the continental scale, these limitations may be less compelling, but at higher spatial resolutions (e.g. at country-wide or regional scales) the accuracy of predictive maps based on such records could elude their intentions [
59,
61]. To overcome some of these limitations, this study assessed habitat suitability for malaria vectors in Cameroon with a species distribution modelling technique based only on presence data gathered from a randomized sampling plan constructed to cover all major bio-geographic domains of the country. Thus, the method used in this study fills a gap in the practical application in both the fields of spatial mapping and statistics and will serve as a stepping-stone for future comparative studies. Areas of improvement in future research will include an ability to employ other presence-only models to compare habitat predicting maps and the accuracy of the predictions [
14,
62].
In Cameroon, the large diversity of malaria vectors (Table
1) is non-randomly distributed across the country (Figure
2). Three different analytical approaches were employed (i.e., ENFA, discriminant analysis, and multivariate regressions) to investigate the environmental requirements and to build optimal habitat profiles of the most common malaria vectors in Cameroon [
3]. One of the most significant results was the high global marginality value found for all the species concerned, indicating that these mosquitoes occupied only a specific set of environmental conditions of those available across the country. This is perhaps not surprising considering that marginality values are related to the extent of the spatial reference set, which in this case was constituted by the whole of Cameroon, a highly diversified country covering several different bio-geographic domains. Moreover, as found in other studies [
32,
63], eco-geographical variables (EGVs) related to human activity (distance to localities and roads) had the most important impact on the ecological niche of anthropophilic malaria vectors. These two EGVs are variables correlated to the density of roads and populated places per unit area, because they take into account the presence of spatial units occupied by localities or roads neighbouring the sampled focal unit. As such, they identify areas where anthropogenic modifications of the environment are greater. This outcome, as well as the anopheline fauna recorded in this study, could be in part related to the collection method, which focused on mosquitoes with domestic resting habits. This is a bias inherent in the fact that the sampling plan targeted those anophelines that are mostly implicated in malaria transmission, which are also those with the most anthropophilic habits (among which the behavioural trait of resting in human dwellings) [
21].
Anopheles gambiae and
An. funestus are highly anthropophilic and endophilic mosquitoes [
33,
64,
65]; sampling bias is, therefore, not expected to have significantly affected the outcome of habitat suitability maps for these species, as confirmed by the excellent predictive performance of the habitat suitability models for these two species (Table
3). Conversely
An. arabiensis is reported to have predominantly exophilic and zoophilic habits in several parts of Africa [
66,
67],
An. nili and
An. moucheti can also be highly exophilic [
19,
68]. Similarly to results obtained by Simard and colleagues [
32], species with less anthropophilic behaviour had a weaker correlation with anthropogenic EGVs (Table
2). This difference in the strength of correlation according to degree of anthropophilic behaviour, suggests that sampling bias due to the collection technique used probably did not invalidate unduly the results of the ENFA and habitat suitability maps. It should be noted, however, that model prediction performance was lower in the case of less anthropophilic species (much so in the case of
An. moucheti), indicating that there is still scope for improvement of the habitat suitability maps produced with the field data available in this study. Nevertheless, similar ecological species distributions were obtained by the ENFA and canonical correspondence analysis. The latter analytical technique takes into account both species presences and absences, and no EGV related to anthropogenic modifications of the environment was introduced as an explanatory variable. Thus, it can be expected that the distribution patterns observed are likely to be of general value for the less anthropophilic species too.
Despite their high global marginality,
An. gambiae and
An. funestus occurred in sympatry in a wide range of ecological settings (Figure
4 and Table
4). Habitat suitability maps predicted large patches of optimal habitat for both species from northern to southern Cameroon (Figure
2A, 2B). However,
An. gambiae was comparatively more associated with conditions characterized by higher rainfall and humidity (Figure
3), which are characteristics of the equatorial rainforest, explaining why the distribution of optimal habitat in this bio-geographic domain was more extensive for this mosquito compared to
An. funestus. It is interesting to note that both species, which showed the lowest global marginality and highest tolerance, have in Cameroon highly structured populations at the genetic level [
69‐
71].
Anopheles gambiae is in fact an assemblage of populations belonging to two molecular forms. Simard and colleagues [
32] analysed the ecological niche requirements of these evolutionarily diverging ecotypes, showing that when considered as separate entities, the marginality and specialization indices were more extreme than those found in the present work for
An. gambiae considered as a single taxonomic entity. Above and beyond the subdivision of
An.
gambiae in molecular forms, this malaria vector exhibits also an extraordinary degree of chromosomal polymorphism, which can be related to its high capacity to adapt to a wide range of ecological conditions [
72,
73]. Similarly,
An. funestus populations in Cameroon are composed of several chromosomal inversion variants with distinct geographical distributions [
69]. The more 'generalist' nature of
An. gambiae and
An. funestus as taxonomic entities, therefore, could well result from the assemblage of natural populations of genetic variants (karyotypes) each having a specialized ecological niche [
63]. Conversely,
An. arabiensis,
An. moucheti and
An.nili had much more restricted and contrasting distributions of suitable habitat (Figure
2C, 2D, 2E).
Anopheles arabiensis was mainly distributed in the most xeric habitats of northern Cameroon that are characterized by high values of evapo-transpiration and sunlight exposure (Table
2). Here, it is frequently found together with the other two major malaria vectors
An. gambiae and
An. funestus, contributing to high rates of parasite transmission during and soon after the rainy season in the savanna bio-geographic domain [
23,
61,
74]. On the other hand,
An. nili and
An. moucheti are two 'forest' species, occurring in regions characterized by higher values of water vapour pressure and rainfall (Table
2), as is typically recorded in the equatorial rainforest of southern Cameroon. Both species, together with
An. gambiae and, to a lower extent,
An. funestus, sustain year-round malaria transmission in the forested regions of Cameroon [
68]. Unfortunately, cytogenetic data are as yet not available for these two species to relate their ecological requirements with chromosomal polymorphism.
This study focused on the role that abiotic variables related to climate, topography, or land use have on malaria vector range and distribution, whereas biological processes such as inter-specific competition or predation were not included among the species distribution modelling predictors. Previous studies have revealed how biological interactions, particularly at the larval stage of development, can affect the population dynamics and distribution of anophelines [
75,
76]. For example, breeding place competition between
An. gambiae and
An. arabiensis can displace the former in favour of the latter [
75]. Contrasting responses to aquatic predators have been considered responsible for generating differences in life-history traits between the two molecular forms of
An. gambiae according to the nature of the breeding site [
65]. The role that these interactions play in the modulation of mosquito geographic distribution and dynamics certainly warrants further research.
Several studies have reported that human disturbance of the natural environment through the action of irrigation or deforestation can favour the spread and colonization of new areas by efficient malaria vectors, increasing the risk of transmission [
29,
64,
77,
78]. In agreement with this view, large patches of optimal/suitable habitat for vectors such as
An. gambiae or
An. funestus occurred in densely populated areas of intensive farming and around major urban centres, whereas regions of low human population density or uninhabited areas were classified as marginal or totally unsuitable. This suggests that global environmental changes, including deforestation, urbanization, or land use conversion for agricultural purposes, as well as the ongoing demographic surge that Africa is currently experiencing are likely to impact on vector distribution and malaria epidemiology in the times to come. Future research should consider the dynamic nature of mosquito population ecology, including population genetic analyses, to understand the evolution of species ranges and current trends in malaria transmission risk.
The presence of a highly differentiated malaria vector system occurring in a given geographical area, as observed in Cameroon, can clearly have a profound impact on the nature and intensity of transmission [
19,
29]. In this context, fine-grained mapping of the vectors' distribution together with the identification, characterization and ranking of their ecological requirements, as well as of the ecological determinants to which mosquitoes respond, is of great interest to assess and predict disease transmission risk. Such knowledge might allow focusing vector control efforts in areas and at times where the target vector species are most amenable to control, and improve both efficacy and cost-effectiveness of disease control through vector control interventions [
62,
79].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
DA, CC, DF and FS conceived and designed the experiments. DA, GCK, CAN, JPA, PAA performed the fieldwork. DA and KO carried out the SIG analysis. DA, CC and FS analysed the data. DA drafted the manuscript, which was critically reviewed by CC and FS. All authors read and approved the final manuscript.