The dominant Anopheles vectors of human malaria in Africa, Europe and the Middle East: occurrence data, distribution maps and bionomic précis

1This paper is a second in a series of three contributions discussing the geographic distribution and bionomics of the dominant vector species (DVS) of human malaria. It deals specifically with the DVS of Africa, Europe and the Middle East.

2Despite highly variable levels of transmission across Africa, the global public heath impact of P. falciparum malaria is overwhelmingly felt on this continent. Africa contains areas with the highest entomological inoculation rates and prevalence levels globally, and thus the highest morbidity and mortality. This situation arises partly because Africa has the most effective and efficient DVS of human malaria: An. gambiae (sensu stricto - herein, referred to as 'An. gambiae'; it is not necessary to use 'sensu stricto' (or the abbreviation 's.s.') when there is no doubt that the biological species being referred to is the one that bears the name An. gambiae), with its sibling, An. arabiensis, also of major importance. The DVS members of the An. gambiae complex also include the salt water tolerant, coastal species An. melas and An. merus and these, whilst not being as efficient at transmitting malaria as An. gambiae or An. arabiensis, are often found in such high densities that they achieve DVS status. Other members of the An. gambiae complex are either highly restricted in their distribution (e.g. An. bwambae, only currently known to occur in geothermal springs in western Uganda) or are zoophilic in behaviour and not considered vectors of human malaria (e.g. An. quadriannulatus and An. quadriannulatus B). In addition to the four DVS within the An. gambiae complex, large parts of Africa are also home to other DVS, including An. funestus, An. nili and An. moucheti, with An. funestus, in some cases, having a greater impact on malaria transmission even than An. gambiae. The anthropophilic habits of these DVS are a major contributing factor to their public health impact, indeed An. funestus is considered to be one of the first species to have adapted to human hosts.

3The vast majority of current malaria control efforts use interventions aimed at limiting human-vector contact. Foremost among these interventions has been the rapid scale-up of insecticide treated bednets (ITNs), followed by the scale-up of indoor residual spraying (IRS) in Africa. These interventions are often deployed without a detailed understanding of the distribution, species composition and behaviour of local vectors. This complicates impact monitoring, the appraisal of arguments for more holistic integrated vector control and evaluation of the potential of novel vector control methods. Distribution maps can also be applied to gauge the importance of emerging insecticide resistance among the DVS of Africa. In contrast to Africa, the European and the Middle Eastern region contain areas with low to no malaria transmission. Despite this, the existence of Anopheles species with the capacity to transmit malaria is often highlighted as providing the potential for the re-introduction of malaria.

4A number of vector species modelling and mapping strategies have been applied on a country (e.g.) and regional scale and across the African continent, with fewer attempts directed at the European and Middle Eastern species. No previous mapping efforts formally incorporate expert opinion (EO) distributions and the methods used range in complexity, from simply plotting presence or abundance on a map, to the application of more sophisticated predictive models. This makes comparison between the maps difficult. Further difficulties also arise in the interpretation of existing maps as many previous studies include all historical occurrence records to compensate for poor data coverage. This can introduce taxonomic ambiguity; the An. gambiae complex, for example, was only fully categorised in 1998, with the addition of the provisionally designated An. quadriannulatus species B and, even now, the status of An. funestus is under question. Moreover, the morphological similarity that hides members of a species complex adds a level of uncertainty to the identity of species data recorded before the advent of cytological or molecular identification techniques.

5This current work attempts to overcome many of these problems. The same Boosted Regression Tree (BRT) methodology is applied to all DVS making comparison between predicted maps possible. Despite only using data collected after 31 December 1984 the assimilated DVS occurrence records together comprise the largest contemporary dataset for prediction, with this evidence base to be made available in the public domain. Significant efforts were also expended to update the EO maps for all species and these were used to inform the predictions. The outcome of these efforts and that of a comprehensive bionomics review are presented here for the DVS of Africa, Europe and the Middle East.

6The data assembly and mapping methods, climatic and environmental variable grid pre- and post-processing methods and the modelling protocol summarised here are described in detail in Sinka et al.. The selection of the DVS is detailed in Hay et al.. In brief, 13 DVS from a final list of 41 species and species complexes worldwide were considered, seven of which are found solely in Africa (Table 1) with a further six distributed across Europe, the Middle East and in limited areas of northern Africa (Table 2).

7Building on the existing Malaria Atlas Project (MAP) library of parasite rate surveys, a systematic search of the published, peer-reviewed literature using online scientific bibliographic databases was performed and augmented with a range of other information previously described. Literature searches were concluded on 31 October 2009 and all citations meeting our search criteria were reviewed.

8Occurrence data extracted from these sources (a detailed protocol is given in Hay et al.) were subjected to a series of rigorous checks before being migrated from Excel into a web-based PostgreSQL database where a final series of checks were conducted (see Sinka et al.).

9Globally, the literature search resulted in 3857 publications or reports containing potential data to be reviewed. Of these publications, 2276 fulfilled the inclusion criteria, providing data for 147 countries. A total of 727 sources detailed surveys conducted across 46 countries in Africa with 45 sources found for 49 countries in Europe and the Middle East.

10Using EO map overlays (Additional file 1: Expert opinion distribution maps for the seven DVS of Africa and the six DVS of the Europe and Middle Eastern region (Raster prediction files are available on request)), initially digitised from published, authoritative sources (Table 1, 2) and further refined by a Technical Advisory Group (TAG) of Anopheles experts (see acknowledgements), preliminary maps were produced displaying the occurrence data for each species. These maps were examined and points that fell outside the EO range were checked and either corrected or the EO maps adjusted to include all confirmed areas of occurrence.

11The BRT method was chosen to generate the predictive maps of each DVS distribution. In a review comparing 16 species modelling methodologies, BRT consistently performed well and benefits from being flexible (accommodating both categorical and continuous data), using freely available, reliable and well documented R code and producing maps that are simple to interpret and include a ranked list of environmental or climatic predictors. The method is described in full by Elith et al. and its implementation for DVS mapping summarised by Sinka et al.. The BRT also produces a number of evaluation statistics including Deviance, Correlation, Discrimination (Area Under the operating characteristic Curve: AUC) and Kappa (κ) which are used here as a guide to the predictive performance of each map.

12The BRT model was provided with a suite of open access, environmental and climatic variable 5 × 5 km resolution grids, relevant to the ecology and bionomics of the DVS in the African, European and Middle Eastern regions. Each grid has undergone a series of processing steps to ensure all land and sea pixels exactly correspond, and, using nearest neighbour interpolation, to fill in any small gaps in the data due to, for example, cloud cover (see Sinka et al.). Where the remotely sensed imagery was available as multi-temporal data, temporal Fourier analysis (TFA) was applied to ordinate the data, generating seven products for each temporal variable: the overall mean, maximum and minimum of the data cycles; the amplitude (maximum variation of the cycle around the mean) and the phase (the timing of the cycle) of the annual and bi-annual cycles. The environmental/climatic variables applied to the BRT model included a digital elevation model (DEM), precipitation and temperature, land surface temperature (LST), middle infrared radiation (MIR) and the normalized difference vegetation index (NDVI) (Advanced Very High Resolution Radiometer (AVHRR)), and 22 individual categories of land cover plus a further three grouped classes that encompassed flooded areas, forested areas and dry areas (Globcover).

13The AVHRR grids (LST, MIR and NDVI) were applied to all DVS except the European species An. messeae and An. atroparvus. These two species have the most northerly distribution of all the DVS, with An. messeae ranging up to 65° north. At these latitudes, the AVHRR satellite data can be problematic. Instead MODIS (MODerate Resolution Imaging Spectroradiometer) data were used because it provides better coverage and fewer data gaps for these northern distributions. The MODIS grids include the Enhanced Vegetation Index (EVI) and LST.

14Following the same protocol described in Sinka et al., numerous model iterations were run to assess the 'optimal' mapping outputs, including assessing the buffer size surrounding the EO range from where pseudo-absences would be drawn, the number of pseudo-absences to apply to the model and the effects of including half weighted pseudo-presence data, allocated at random from within the EO boundary, alongside the occurrence data. As each of these categories required the use of different data inputs to the BRT, statistical comparison using the evaluation metrics was not strictly possible. Therefore the 'optimal' settings chosen are inherently subjective and based on visual examination and comparison of the various maps guided by, but not relying on, the evaluation statistics.

15A full protocol describing the methodology used to extract species-specific bionomic data from the available literature (Table 3, 4) is given in the supplemental information accompanying Sinka et al.. The bionomics summary of each species is included to accompany the predictive maps as the success of interventions and control methods, such as ITNs or IRS, in reducing malaria transmission is closely related to the behavioural characteristics of the local DVS. This review does not, however, include detailed information relating to insecticide resistance. This was a purposeful omission as it would not be possible to do full justice to this highly dynamic and important aspect of the DVS within the space confines of the current work. Moreover, insecticide resistance is being addressed in detail by other groups, including those at the Liverpool School of Tropical Medicine and the Innovative Vector Control Consortium (IVCC). Furthermore, there are a number of comprehensive reviews that have been recently produced that detail insecticide resistance amongst Afrotropical species which should be considered alongside this current work (e.g.).

16A total of 4581 independent sites, of which 4234 were successfully geo-referenced, reported the presence of one or more African DVS, relating to 9300 (8646 geo-referenced) occurrences (i.e. including one or more temporal sample conducted at one independent site) (Table 5). The following results refer only to geo-referenced data, and of these 3951 sites were at a resolution (points and wide areas, <10 km2 and between 10 and 25 km2 respectively) suitable to be applied to the BRT model (from here on, for simplicity, referred to as points).

17Data were recorded from a total of 46 countries, 44 of which reported points. The largest number of data were reported from Kenya, with a total of 757 sites (all area types), 686 points and 1599 occurrence data (all area types). In contrast, only one data point was reported from Togo (Kantindi) where An. gambiae was found and studies from Mauritius only provided DVS location information, at a polygon level, for two sites. African DVS data were reported from Egypt, but only in the form of a polygon location that could not be successfully geo-referenced. Anopheles gambiae was reported from the largest number of countries (34) and from the highest number of point locations (1443), however occurrence data (from point locations only) were greater for both An. funestus and An. arabiensis (2692 and 2301, respectively) than for An. gambiae (2291). The least prevalent species was An. moucheti reported from only 66 point locations (Table 6) and Cameroon had the highest diversity of DVS with three sites (Nkoteng, Tibati and Mayo Mbocki) showing the presence of five DVS (An. arabiensis, An. funestus, An. gambiae, An. nili and An. moucheti).

18Adult resting collections were the most popular sampling method, with 424 studies collecting females resting inside houses compared to 178 studies that collected females biting indoors. Outdoor resting sampling was comparably rare with 56 studies collecting from outdoor shelters, 22 studies searching inside animal sheds and 21 studies where the details of the outdoor location sampled were not recorded. Outdoor landing catches were conducted in 132 studies and 181 studies collected larvae, relating to 675 point locations.

19Molecular techniques examining nucleic acids, which have only been applied for identification on a regular basis since the 1990s, were well represented, with 338 studies reporting the use of Polymerase Chain Reaction (PCR) methods. Morphological methods were used in 363 studies, often in conjunction with PCR techniques. At the other end of the scale, salinity tolerance tests were only attempted in four studies and cross-mating experiments only in five.

20Across the European and Middle Eastern region, 49 countries reported the presence of one or more DVS from 2820 point locations (all locations: 2891), of which 2784 were successfully geo-referenced (all geo-referenced locations: 2848) (Table 7). Relatively few polygon data were reported (all: 71/2891, georeferenced only: 64/2848) and longitudinal studies were also rare, with only 18 studies reporting sampling on more than one occasion at the same site. A total of 3020 geo-referenced occurrence data across all area types, with 2946 from point locations, were compiled. Considering only the geo-referenced data, DVS presence was reported from the most sites in Italy (all sites: 423, point only: 409).

21Anopheles atroparvus was the species reported most often across the region, found at 1051 geo-referenced locations, of which 1044 were available to be used in the analyses. Anopheles sergentii was only present at 35 point locations, and within 11 polygon areas, but these related to a total of 113 occurrence data (102 points, 11 polygons) (Table 8).

22In the European and Middle Eastern region, larval collections were the most common sampling method, with 23 studies sampling at 86 sites. Sampling methods were unknown for a large proportion of the data (1553 sites), of which 1488 related to a single data source. Possibly due to the zoophilic nature of the majority of the European and Middle Eastern species (see below), resting adult females were collected from animal sheds at 85 locations compared to only 31 where resting collections were conducted inside human dwellings. Human landing collections were conducted indoors in only two studies, relating to only three sites, with three studies collecting by outdoor human landing at only eight sites.

23Identification methods, amongst those studies that reported them, mainly relied on morphological characteristics and were conducted on specimens from 175 locations. Only 10 studies reported using PCR identification techniques but due to a large number of unknown or unreported methods, this ranked as the second most popular method, and was applied to specimens collected from 67 sites.

24The results for each mapping trial are given in Additional file 2 (Additional file 2: Summary tables showing evaluation statistics for all mapping trials and final BRT environmental and climatic variable selections for the final, optimal predictive maps). Optimal mapping categories were evaluated visually and using the deviance and AUC statistics, with the caveat that these could only be used as a guide rather than a definitive indication of predictive performance.

25The EO mapping test indicated that where random pseudo-presences were created within the EO range, and no real occurrence data were included, the model would predict a high probability of presence within the whole EO range and calculate a high deviance value for all species, indicating an overall poor predictive performance. This was the case for the African species and those from the European and Middle Eastern region, and consistent with the results for the nine DVS in the Americas. Where the hybrid method was used that incorporated both real occurrence data plus 500 half-weighted pseudo-presence points randomly assigned within the EO range, the mapping performance was greatly improved. Maps created using only the real presence data produced a low deviance value, but visually, predictive performance was judged to be poor, possibly due to a paucity of data for some species. It was therefore considered that the hybrid maps performed better overall and are presented here.

26The optimal buffer width for the African DVS was judged to be 1500 km, producing the lowest deviance value for five out of the seven species. For the European and Middle Eastern species maps, all buffer widths other than 1000 km had high deviance values for all species. The 1000 km buffer therefore was judged to perform better for all six species and applied consistently to all final maps.

27For both the African and the European and Middle Eastern species, a ratio of 10:1 pseudo-absences to presence data (not taking into account the 500, half weighted pseudo-presence created in the hybrid maps) was judged to perform better overall, but for both regions, the number of pseudo-absences appeared to have little effect on the predictive maps.

28The BRT maps for all seven African DVS and for the six European and Middle Eastern species are given in Additional file 3 (Additional file 3: Predictive species distribution maps for the seven DVS of Africa and the six DVS of the Europe and Middle Eastern region). Spatial constraints prevent all species being discussed in detail here, however, Anopheles gambiae (Figure 1) is the iconic and possibly the most important vector of malaria, and therefore is discussed further below.

29There have been a number of attempts to model the distribution of An. gambiae but the majority tend to focus on single countries and often just map presence points or abundance without further analysis (e.g.). Continent-wide predictive maps for An. gambiae (plus other members of the An. gambiae complex) have also been attempted, making use of satellite-derived environmental or climatic variables (Table 9). The methods range from simply overlaying presence and absence points over rainfall maps to the application of more complex, spatial ecological niche models.

30Precipitation, in one form or another, is identified repeatedly in previous models (where these data are presented, Table 9) as an influential variable in predicting the range of An. gambiae. Within the top five contributing covariates from the suite applied to the BRT model, precipitation was identified three times, with mean precipitation as the highest contributor with a relative influence of over 37%. Maximum precipitation was placed second (19.42%) with the amplitude of the bi-annual cycle of precipitation ranked forth (8.85%). In common with the Maxent niche model presented by Moffett et al., elevation (altitude) and minimum land surface temperature were also identified by the BRT model within the top five influencing climatic/environmental variables (relative influence of 12.36% and 5.68%, respectively).

31Anopheles gambiae larvae are commonly found in temporary, shallow, small bodies of water, such as puddles in hoof prints, wheel ruts and small ground pools (see below), sites which are only present after rainfall. Hence the high influence of precipitation on the distribution of this species identified by the BRT model corresponds closely with the known bionomics of An. gambiae.

32The predictive map of An. gambiae (Figure 1) loosely follows the boundary and distribution indicated by the EO map (Figure 1, inset) with one clear exception: the large gap in the range over southern Kenya and a large proportion of northern and central Tanzania. This gap may be driven by the presence of savannah-type vegetation more commonly associated with An. arabiensis, or the increasing altitude of this region, and may be causal to the identification of elevation as an influencing factor to the distribution of An. gambiae. Similar gaps are also seen in the maps produced by Moffett et al. and, to a slightly lesser extent, in the map of Levine et al.. In Madagascar, Léong Pock Tsy et al. identified altitude as a limiting factor for An. gambiae with numbers diminishing as altitude increased until, other than two specimens found at 1300 m, it was considered essentially absent over 1000 m. However, in the Kenyan highlands An. gambiae is commonly identified up to 2000 m and specimens have been confirmed at sites up to 1800 m in Uganda. Sampling across Africa, as stated by Coetzee, reflects the distribution of entomologists and not necessarily the distribution of the mosquitoes, and the area within this predicted gap, along with a great swath through central Africa, is clearly lacking in empirical occurrence data. Acknowledging these caveats, and similar ones in parts of the range of many of the DVS, it is obvious that samples from these poorly known areas would help improve substantially our predictive mapping.

33Anopheles arabiensis, when compared to An. gambiae, is described as a zoophilic, exophagic and exophilic species. However, it is also known to have a wide range of feeding and resting patterns, depending on geographical location. This behavioural plasticity allows An. arabiensis to adapt quickly to counter indoor IRS control, where suitable genotypes occur, showing behavioural 'avoidance' (deterrence from a sprayed surface) depending on the type of insecticide used.

34Anopheles arabiensis is considered a species of dry, savannah environments and sparse woodland, yet it is known to occur in forested areas, but only where there is a history of recent land disturbance or clearance. Its larval habitats are similar to those of An. gambiae (see below): generally small, temporary, sunlit, clear and shallow fresh water pools (Table 10), although An. arabiensis is able to utilize a greater variety of locations than An. gambiae, including slow flowing, partially shaded streams and a variety of large and small natural and man-made habitats (Tables 11, 12). It has been found in turbid waters and, on occasion, in brackish habitats (Harbach, unpub. obs.). It readily makes use of irrigated rice fields (Table 11), where larval densities are related to the height of the rice, peaking when the plants are still relatively short and then dropping off substantially as the rice plants mature. Such density fluctuations are also reflected in the adult population, which also peak when rice stalks are small and decline as the plants mature. These patterns may be due to a preference for sunlit areas of water with relatively limited emergent vegetation (Table 10), with densities decreasing as shade from the growing plants increases. Moreover, there is evidence that An. arabiensis may be attracted by the application of fertilisers or by the amount of dissolved oxygen within the paddy water. However, with fertiliser application occurring at the start of plant cultivation, and dissolved oxygen content related to sunlight exposure (e.g. via increasing photosynthesis), the primary oviposition attractant in rice fields is uncertain.

35The behavioural variability of An. arabiensis is clearly evident (Table 13), with similar numbers of studies reporting either anthropophilic or zoophilic behaviour. Bøgh et al. stated: 'There is... great variation in the feeding preference depending on the local variation in host availability and composition of the local genotypes of the vector'. Tirados et al. suggested the existence of an east-west behavioural cline. They proposed that those populations found in western Africa display higher levels of anthropophily, and preferentially feed and rest indoors, whereas those in the east exhibit greater zoophily and rest outdoors. Overall, however, biting patterns tend to be exophagic, but such behaviour is often reported in comparison with highly endophagic species such as An. gambiae. For example, Fontenille et al. reported An. arabiensis as 'more exophagic than An. gambiae and An. funestus' with 65.4% of vectors found biting outdoors identified as An. arabiensis, yet 59% of those found biting indoors were also identified as An. arabiensis.

36Blood feeding times also vary in frequency but biting generally occurs during the night. Peak evening biting times can begin in the early evening (19:00) or early morning (03:00) (Table 13). This species does, however, demonstrate a predisposition to exophilic (or partial exophilic) behaviour regardless of where it has blood fed or the source of its meal, a behavioural trait considered to be related to polymorphic chromosomal inversions, to a greater or lesser extent, depending on location.

37Anopheles funestus is a member of the Funestus Subgroup (often mistakenly referred to as An. funestus complex), which includes: An. aruni, An. confusus, An. funestus, An. parensis and An. vaneedeni. The members of this subgroup exhibit important variation in their biology and behaviour, especially in regard to malaria vectorial capacity and are only morphologically distinguishable during certain stages in their development. Only An. funestus is regarded as an important vector of malaria in this subgroup.

38A typical An. funestus larval habitat is a large, permanent or semi-permanent body of fresh water with emergent vegetation, such as swamps, large ponds and lake edges. Larvae have been found in shaded and sunlit environments (Table 10) and Gillies & de Meillon concluded that An. funestus uses emergent vegetation as refuge against predation while the shading it casts, or the presence of shade from overhanging plants, is of lesser importance. In some areas, An. funestus larvae, as with An. arabiensis, are associated with rice cultivation (e.g. Madagascar, Mali) (Table 11). Where they are found, their favoured environmental conditions are very different to those of An. arabiensis. Anopheles funestus replaces An. arabiensis in a successive temporal process during rice plant growth, exhibiting higher densities in older, maturing fields compared to the preceding open conditions preferred by An. arabiensis.

39Anopheles funestus is considered to be highly anthropophilic (Table 13) (but see below), which led Charlwood et al. to propose that An. funestus may have been the first anopheline species to specialise on biting humans, surmising that its preferred larval sites (permanent water bodies in savannah-like environments) are likely to have been areas where humans first settled. Behaviourally, its late-night biting patterns would also allow ready access to human blood without incurring undue density-dependant host avoidance. This late-night biting preference is clearly evident throughout its range, with all studies reviewed reporting a peak biting period occurring after 22:00, and most commonly between midnight and the early hours of the morning (Table 13). Endophilic resting behaviour is also commonly reported, and combined with a relatively high longevity, makes it as good a vector, or better in some areas, as An. gambiae. These characteristics are also responsible for promoting the success of vector control using IRS and ITNs. However, this exposure has resulted in selection pressure and rapid development of insecticide resistance to pyrethroids, now well established in some populations and implicated as the primary reason for a major resurgence of epidemic malaria reported in Kwazulu-Natal, South Africa in the late 1990s.

40Compared to other DVS in Africa, An. funestus shows fairly consistent behaviour (generally anthropophilic and endophilic) throughout its range; however, it is a highly adaptable species, allowing it to occupy and maintain its wide distribution and utilise and conform to the many habitat types and climatic conditions contained therein. Behavioural differences between chromosomal forms have been identified, for example, Lochouarn et al. reported anthropophilic behaviour in western Senegal and zoophilic behaviour in the east of the country, behaviours which correspond to chromosomal polymorphisms that also follow this east-west cline. Costantini et al. identified two chromosomal forms in Burkina Faso associated with different resting and biting behaviour. This, coupled with a lack of heterokaryotypes in areas where the two forms co-exist, prompted these authors to suggest that the two forms were incipient species, and hence of the concept of an An. funestus complex. More recently, An. funestus populations from 12 countries have been divided into three molecular types: M, W, and MW, correlating to geographical locations, whereby M is essentially found in eastern Africa, W from western and central Africa, and MW from southern Africa. Further investigations showed a more complicated situation with specimens from Malawi showing all three types, specimens from Tanzania showed the M- and MW-types, whereas specimens from Kenya showed M- and W-types. In addition, two more types were described, type Y from Malawi, and type Z from four localities of Angola, Malawi, Ghana and Zambia. Finally, adding further to the complexity surrounding this species, recent studies in Malawi have revealed a new species of the subgroup, named An. funestus-like that is identical to An. funestus but appears to have a different biology and role in malaria transmission, although this needs confirmation.

41Anopheles gambiae is considered to be one of the most efficient vectors of malaria in the world and is one of the most well studied. Like An. funestus, the variable ecological conditions present within the large geographical range of An. gambiae indicate a highly plastic species with corresponding chromosomal diversity currently separated into five chromosomal forms: Forest, Bamako, Savanna, Mopti and Bissau. There is suggestion of reproductive isolation among the sympatric forms, and hence, of incipient speciation between them. Independent of these chromosomal categories, two molecular forms, 'M' and 'S', have also been described, and are the forms more commonly referred to in the recent literature. These different forms exhibit ecological adaptations which further indicate possible speciation, for example the Mopti and M forms are associated with semi-permanent, often man-made, larval habitats such as rice fields or flooded areas, whereas the Savanna/Bamako and S forms are seen more commonly in temporary, rain-dependent sites such as ground puddles. There appear to be no definitive studies that explicitly describe variability in adult biting or resting behaviour or role in malaria transmission between the two molecular forms.

42Despite its wide range and variable ecology, a combination of traits allows An. gambiae to maintain its position as one of the most efficient vectors in sub-Saharan Africa. It is a relatively long-lived species (although not as long as An. funestus), with a short larval development period and is often found in larval habitats associated with human activity (e.g. water in hoof prints, wheel ruts or areas of rice cultivation) (Tables 11, 12). It is considered to be highly anthropophilic, with 11 of 15 studies that examined biting behaviour (Table 13) reporting a marked preference for human hosts. However, there are a number of studies that indicate An. gambiae is less discriminant and more opportunistic in its host selection and that host choice is, as with the majority of African DVS, highly influenced by location, host availability and the genetic make-up of the mosquito population. Moreover, many studies that report host preference using blood meal analysis are often conducted on resting, blood-fed specimens collected inside houses, thus introducing a potential study design or sampling bias favouring the likelihood that the blood meal will be from a human host. Of the studies that report some level of zoophily, Diatta et al. specifically examined the host preference of An. gambiae and An. arabiensis by comparing the number of females of each species captured either in a calf-baited or a human-baited net trap. There was no statistical difference between the host preferences of the two species, both expressing greater zoophily (e.g. 31% of An. gambiae were found in the human-baited trap and 69% in the calf-baited trap). Duchemin et al. also reported zoophilic behaviour, yet highlighted this as unusual, suggesting that the high density of cattle in the sampling area may have influenced the propensity for zoophily in the population. Bøgh et al. reported no specific preference for either human or animal hosts but that An. gambiae would feed readily on cattle.

43As with An. arabiensis, An. gambiae larvae typically inhabit sunlit, shallow, temporary bodies of fresh water such as ground depressions, puddles, pools and hoof prints (although see above) (Table 10, 12). Gillies & de Meillon suggested that this aspect of their bionomics allow members of the An. gambiae complex to avoid most predators, and the larvae are able to develop very quickly (~six days from egg to adult under optimal conditions and temperatures), possibly in response to the ephemeral nature of their larval habitats. Water in these larval sites can appear clear, turbid or polluted (Table 10). Typically An. gambiae larval habitats are described as containing no (or very sparse) vegetation (Mbogo, unpub. obs.) due to their temporary nature. Gillies & de Meillon summarised the great diversity of habitats utilised by An. gambiae, and as described before, different molecular or chromosomal forms are associated with either vegetated (e.g. rice fields) or temporary and non-vegetated (e.g. hoof prints) larval sites. The studies reviewed here report An. gambiae from habitats containing floating and submerged algae, emergent grass, rice, or 'short plants' in roadside ditches and from sites devoid of any vegetation (Table 10).

44Females of An. gambiae typically feed late at night, a characteristic shared with An. funestus that may increase their ability to effectively transmit malaria parasites (see above) (Table 13). Anopheles gambiae is often described as an endophagic and endophilic species, both biting and resting indoors, however, the majority of studies listed herein (nine of 11), that compared indoor and outdoor human-landing catches reported no difference in the numbers of females collected at either location and an equal number of studies recorded post-feeding exophilic resting as resting indoors. Bockarie et al. linked differences in the exo-or endophilic behaviour of An. gambiae to their chromosomal forms, suggesting the Forest form (with no inversion) demonstrated stronger exophily in southern Sierra Leone whereas the Savannah form, with a 2La inversion, was mostly endophilic. Odiere et al. used clay pots to sample outdoor resting females in western Kenya and found no clear preference for indoor or outdoor resting. They suggested that the designation of An. gambiae as a predominantly endophilic species may have been based on poor sampling comparisons. As with host preference, this species appears to exhibit greater phenotypic plasticity and opportunism in blood feeding and resting locations than commonly thought.

45There is relatively little contemporary information about the behaviour of An. melas, perhaps because it is generally considered to be a vector of lesser importance, specifically where it occurs in sympatry with An. gambiae or An. arabiensis. Anopheles melas has a comparably lower sporozoite rate than either An. arabiensis or An. gambiae (e.g. 0.35% compared to 3.5% for An. gambiae in The Gambia), yet in coastal areas where it can occur in very high densities it is still a problematic vector of malaria. With the dearth of available contemporary data, those studies conducted prior to 1985 that closely examined the behaviour of this species have been included here.

46Anopheles melas is commonly associated with brackish water and can utilise saline environments that other species, for example, An. gambiae, cannot tolerate, yet does not appear to require brackish water for larval stage development. It is generally restricted to coastal areas but has been found up to 150 km inland along the Gambia River, where salt water can intrude great distances (up to 180 km) upriver. Unlike other African DVS, the density fluctuations of An. melas are closely associated with tidal changes rather than seasons, for example, Gelfand identified a peak in adult numbers 11 days after spring tides. The larvae of this species are associated with salt marsh grass (Paspalum spp.) and mangroves, but only trees of the genus Avicenna, which include white, grey and black mangrove, and not those from the genus Rhizophora ('true' or red mangrove spp.). These positive and negative associations with mangroves are thought to be strongly influenced by the predominant soil type associated with the different tree genera. Anopheles melas preferentially oviposits on damp ground at low tide, rather than in open water, where the eggs are able to survive some degree of desiccation until the tides rise again, and appears to prefer the poorly drained, peaty-like soil common to Avicenna forests compared to the sandy, gravelly or smooth, fibrous peat soils common to the Rhizophora stands. Giglioli surmised, that this behaviour guarantees the larvae will have sufficient time to complete their larval development and pupate in the less saline, relatively permanent waters of the new tide before it begins to recede and the water either becomes too salty, or dries out completely.

47Adult biting behaviour appears to be opportunistic. Anopheles melas has been described as both highly anthropophilic and a zoophilic species. In a choice experiment, Muirhead-Thomson varied the numbers of animal and human baits in traps to attempt to describe host preference. He found An. melas to be fairly indiscriminate: where there were more animal baits, An. melas would feed more often on animals, but still feed on humans. On the contrary, where there was an increase in the number of human hosts, a sharp decrease in the number of females feeding on animals occurred. Sampling bias towards anthropophily may be reported when blood fed females collected resting inside houses are tested for host blood type because An. melas generally appears to rest outdoors after feeding, although there has been limited success in locating and collecting from such natural outdoor resting sites. As previously described for An. gambiae, those females that bite and rest indoors are more likely to have fed on humans, and those biting or resting outdoors (or in animal sheds) are more likely to have bitten animals. Blood feeding activity appears to be fairly continuous throughout the night. Gefland observed continual biting from 19:00 to dawn, although Muirhead-Thomson saw two peaks of biting activity: the first, and slightly smaller peak, between midnight and 02:00 and a second, larger peak, between 04:00 and dawn.

48Anopheles merus has previously been considered as only a minor, or even an unimportant vector, potentially unable to sustain malaria transmission alone. However, is has been identified as playing an 'unexpectedly important role' along the Tanzanian coast and more recently in Mozambique. It is also a species for which there is limited contemporary information. The differences in egg and larval morphology that distinguish An. melas from An. gambiae do not occur in An. merus and identification, before the advent of molecular techniques, was based on physiological characteristics involving larval salinity tolerance tests. Originally, An. merus was referred to as a 'salt water An. gambiae' variant or subspecies. Indeed, Jepson et al. had a number of specimens collected in the 1940s from saline, coastal swamps in Mauritius examined for distinguishing features, and found no obvious morphological distinguishing characters and stated 'All the specimens proved to be typical forms [of An. gambiae] and there was no evidence of the presence of An. gambiae var. melas'. They continued to regard 'An. gambiae' as a species with 'a considerable tolerance for pollution and salinity and is therefore to be found in domestic wastes and in crab holes and pools near the sea side, in addition to a host of natural breeding places such as marshes, rock pools and casual rainwater pools'. This Mauritian species was finally designated a subspecies of An. gambiae by Halcrow, who provisionally named it An. gambiae litoralis based on larvae found in '...water of high salinity in crab holes, depressions in coralline rocks, small tidal lagoons, pools close to tidal zone and [interior] salt pans, and are not associated with mangroves...'. Paterson provided definitive proof of the specific status of An. merus and the validity of the name.

49Halcrow's description highlights a specific difference between An. merus and An. melas. Anopheles merus is rarely found in the mangrove forests on the east coast, however this may be due to the composition of the trees and soil type under of the stands of mangrove in this zone rather than inherent behavioural differences between the two species. Anopheles merus is, instead, found in high numbers in shallow brackish pools and marsh or swamp areas along the coast. As a consequence, this species does not exhibit density changes in response to the tidal fluctuations as seen with An. melas, nor does it appear to tolerate the same high levels of salinity. Anopheles merus is also known to occur further inland, using salt pans and saline pools larval habitats, and cross-mating experiments between inland and coastal populations have produced viable offspring indicating they are conspecific.

50The biting behaviour of An. merus is similar to that of An. melas: generally opportunistic in host selection, depending on host availability and with a tendency to bite and rest outdoors. Gillies & de Meillon suggested that An. merus shows a preference for animal hosts, referring to a laboratory test where, given a choice, females consistently fed on calf versus human bait. Two of the studies reviewed here reported anthropophily, one indicated zoophily and another concluded that no obvious preference was detected. In the latter study, blood meal analysis was conducted on mosquitoes collected resting indoors (59.2% had fed on humans), and those collected resting outdoors (71.4% had fed on cattle and only 1.6% contained human blood), highlighting the bias in drawing conclusions on host preference if only indoor or outdoor resting specimens are tested. Only one study, conducted on the Kenyan coast, examined the biting times of An. merus, which reported the number of bites gradually rising from early evening (18:00) peaking between midnight and 01:00 and then declining to 06:00 which corresponds to the accepted biting pattern for this species across its range (Bangs and Mbogo, unpub. obs.).

51Anopheles moucheti is a species with two morphological forms: An. moucheti moucheti, and An. m. nigeriensis which are distinguishable by morphological features of the adult and larval stages. Anopheles m. bervoetsi, previously considered a third morphological form, has recently been raised to full species status: An. bervoetsi by Antonio-Nkondjio et al.. However, these authors do assert a level of caution in this new status as they point out that An. bervoetsi has only ever been reported from its type locality (Tsakalakuku, DRC) and has never been found in sympatry with An. moucheti. They do cite unpublished data that detected P. falciparum infection in An. bervoetsi specimens, and thus raises the possibility that this species could be transmitting malaria in central Africa. The bionomic information detailed here is, in the most part, taken from sources that present data for 'An. moucheti'. Of these, the majority of studies have been conducted in Cameroon by Antonio-Nkondjio and colleagues or in Nigeria, so based on current knowledge the assumption is that these data refer to An. moucheti and not An. bervoetsi.

52Despite its status as a DVS, An. moucheti is a poorly studied species. It is the only DVS with its range entirely restricted to forested areas, specifically where the canopy is broken allowing sunlight to penetrate to the ground, such as is found where large rivers flow through the forest. Human activity, such as road building, settlements or cultivation, can therefore be beneficial to this species by breaking up the forest canopy, although larger areas of deforestation may decrease the density of An. moucheti and allow replacement by An. gambiae. Anopheles moucheti larvae are found at the edges of large, slow flowing or lentic rivers, often with turbid waters, and are associated with Pistia spp (water lettuce/water cabbage). Antonio-Nkondjio et al. studied the larval habitats along the river networks of southern Cameroon and found the greatest numbers of An. moucheti larvae along the margins of rivers within deep, evergreen forest, substantially fewer in the degraded forest and none in the savannah areas. Where they were found, larvae were abundant near to areas of human habitation.

53Although the range of An. moucheti is relatively restricted within the equatorial forests, it derives its status as a DVS from its highly anthropophilic and endophilic behaviour (Table 13). Gillies & de Meillon suggested such behaviour is unsurprising due to the lack of domestic animals found within forested environments. Anopheles moucheti is also described as highly endophagic, however this characteristic appears to be less than clear cut. For example, Antonio-Nkondjio et al. found that in urbanised, forested environments (where An. moucheti was less abundant and replaced by An. gambiae) compared to rural localities (where An. moucheti was dominant), only 43% of females were found biting indoors, whereas in the rural areas 66% were found biting indoors. In a study conducted in a village only 2 km from Yaounde, Cameroon, Antonio-Nkondjio et al. reported 51% biting indoors and described the sampled populations as 'mainly endophagic'. Overall, An. moucheti appears endophilic (Table 13). In a countrywide survey of Cameroon, of all females found resting, 1234 were located indoors, whereas only 12 were captured in outdoor shelters. Only two studies examined the biting cycle of An. moucheti, with both reporting biting gradually increasing towards the second half of the night to dawn; Mattingly reported peak biting activity in the early morning between 03:15 and 06:15.

54The An. nili complex includes An. carnevalei, An. nili, An. ovengensis and An. somalicus. As with An. moucheti, species of this complex have been generally overlooked in African vector studies despite being described as highly efficient vectors. Amongst members of the complex, An. nili is considered the most important vector, although An. carnevalei and An. ovengensis are implicated as secondary vectors of P. falciparum in Cameroon. Anopheles somalicus is considered zoo- and exophilic: it was not found to bite humans in Somalia and no females were found in houses in Cameroon despite an abundance of larvae in the area.

55Larvae of all members of the An. nili complex are found in vegetation at the edges of fast flowing streams and rivers. However, An. ovengensis and An. carnevalei appear to be restricted to areas of deep forest, whereas An. nili is more abundant along rivers in degraded forest and savannah. A comprehensive survey of the river systems across Cameroon found An. nili larvae associated with sunlit sites whereas An. carnevalei larvae were more commonly found in shaded areas.

56Anopheles nili is considered to be strongly anthropophilic, and will readily bite both indoors and out (Table 13). Carnevale & Zoulani described biting patterns that exploited the behaviour of their human hosts, biting outdoors in the early evening when people are socialising, and then continuing to bite indoors once people move inside, with peak feeding occurring after midnight. The resting habits of An. nili are also described as 'variable'. Krafsur, in a lowland region of western Ethiopia, rarely found An. nili resting indoors despite the high densities found biting indoors, indicative of exophilic behaviour. Conversely, Antonio-Nkondjio et al. examined populations across Cameroon and reported An. nili overwhelmingly resting indoors (466 females), with only one female captured in an outdoor shelter. In the same study they found no An. carnevalei females resting indoors or in outdoor shelters whereas all resting An. ovengensis captured were found indoors. Conversely, Awono-Ambene et al. stated that An. ovengensis was rarely found resting indoors and concluded it had 'exophilic habits'.

57Anopheles atroparvus is a member of the Maculipennis Subgroup, which also includes An. (Ano.) daciae, An. (Ano.) labranchiae, An. (Ano.) maculipennis, An. (Ano.) martinius, An. (Ano.) melanoon, An. (Ano.) messeae, An. (Ano.) persiensis and An. (Ano.) sacharovi. Of these, An. labranchiae, An. messeae and An. sacharovi are also designated as DVS (see below).

58Anopheles atroparvus is described as a species with a preference for brackish larval habitats. Hackett & Missiroli summarised: 'In general it may be said that over its extensive range [An.] atroparvus is found in water of moderate salinity not exceeding 10 parts per 1000. It prefers relatively cool water, and its range does not overlap that of [An.] labranchiae, a warm water breeder'. However, the larval sites listed in the literature still include a number of predominantly fresh water habitats, for example canals, ditches, river margins, pools in river beds and rice fields, and Cambournac defines An. atroparvus as a 'fresh water breeder'. Hackett also stated that, in southern Europe, An. atroparvus 'inclines to breed in fresh water'. Of the few studies reporting primary data (Tables 14-17), larvae were identified in marshes and ditches/ground flood pools, pools in river beds, river margins and streams, rock pools, cement tanks, rice fields, wells and ground pools and in small collections of water in used tyres (Tables 15, 16).

59Becker et al. described sites to be 'usually sun exposed' and to contain 'a considerable amount of filamentous green algae and other floating submerged vegetation'. Pires et al., in a study that sampled comprehensively across Portugal, reported An. atroparvus larvae to be found more frequently in sun-exposed habitats, although 'some shade was provided by grasses and aquatic vegetation'. They also reported filamentous algae present in 48 of 93 sites positive for An. atroparvus (Table 14).

60Anopheles atroparvus is generally considered zoophilic, and described as 'very zoophilic' by Cambournac, who also stated that its hosts, in order of preference, are rabbit, horse, cow, pig and sheep, and suggested that a long association between rabbit and An. atroparvus (since approx. 1000 BC) may be responsible for this hierarchy of preference. Indeed, An. atroparvus has been implicated as an effective vector of the myxomatosis virus to domestic rabbits in the UK. Elsewhere, however, An. atroparvus is described as anthropophilic, which perhaps indicates the opportunistic nature of this species. Four studies identify An. atroparvus as zoophilic and one study, that did not distinguish a preference, reported the collection of An. atroparvus during night catches on horse bait, from indoor resting sites and during day- or night-time catches on humans (Table 17). There is no clear evidence or information among any of the published studies, nor within the general literature, that identifies this species as preferentially biting indoors or outdoors. The opportunistic nature of its feeding habits and zoophilic proclivity in host choice, however, would suggest it is probably exophagic but that biting location could also depend upon the setting and accessibility of the host.

61Anopheles atroparvus rests and hibernates in animal sheds and stables. It hibernates as an adult female and is known to periodically feed, specifically if she has taken refuge in a relatively warm locality, but these meals do not result in egg production (i.e. gonotrophic disassociation).

62A number of investigators have discussed the inability of An. atroparvus to transmit tropical strains of P. falciparum, with most referring to studies conducted by Shute. Unfortunately this reference could not be found, but in a study testing the susceptibility of Russian anopheline species to imported P. falciparum, no infection was detected in An. atroparvus females. Curtis & White concluded (also referring to Shute) that An. atroparvus is refractory to both Asian and African P. falciparum but competent in supporting a European strain, a conclusion reiterated by de Zulueta et al. with Cambournac stating that refractoriness of An. atroparvus to African and eastern strains of P. falciparum is an 'uncontroversial fact'. However, Capinha et al. claimed the existence of local An. atroparvus in Portugal that could be infected with 'exotic strains of plasmodia', with reference to a comprehensive study by Souza. However, on closer examination of these findings, even though Sousa did indeed infect An. atroparvus with P. falciparum, this was only after numerous attempts that resulted in formation of oocysts in only five out of 736 females. It would seem, therefore, that although An. atroparvus can be infected by tropical P. falciparum strains, it is very unlikely to happen under natural conditions and there is currently no conclusive evidence that such infection would result in salivary gland invasion by sporozoites.

63Despite similarity in larval site characteristics, An. labranchiae and An. atroparvus do not, or only have limited, overlap in their distributions. This lack of sympatry may be simply a factor of temperature, with An. labranchiae making use of warmer waters than typical of An. atroparvus. However, when Capinha et al. modelled the habitat suitability of An. atroparvus across Portugal, they concluded that the most suitable locations include drier areas with higher temperatures (i.e. conditions where An. labranchiae typically dominate), whereas wetter areas with milder temperatures, where An. atroparvus are mostly found, were unsuitable. They concluded that An. atroparvus is not found in many other 'suitable' Mediterranean areas due to competitive exclusion. Conversely, de Zulueta suggested that the absence of An. atroparvus in Sardinia allowed the wide distribution of An. labranchiae on the island, where, despite a five-year eradication campaign instigated in 1946, An. labranchiae still occurs.

64Both species utilise brackish water marshes and lagoons along the coast, although in contrast to An. atroparvus, An. labranchiae will preferentially oviposit in fresh water. Marchi & Munstermann, in a survey conducted across Sardinia, only identified An. labranchiae in fresh water sites, including rock holes, pits, ditches, drains or canals, streams/rivers, flooded ground pools and ponds, lakes or reservoirs. Despite an ability to tolerate some salinity, An. labranchiae larvae are not generally found at sites with significant levels of organic or mineral pollutants (, Mouchet, pers. com.). Larval sites are typically described as sunlit, although in Sardinia Aitken found larvae in 'almost every type of habitat except the very densely shaded', and Macdonald also associated this species with habitats that have some level of shade. In general, An. labranchiae larvae are found in stagnant or slow moving waters and can make use of, and become very abundant in, rice fields. Indeed, Bettini et al. described a survey in central Italy that identified high numbers of larvae in newly established (two years old) rice fields with correspondingly high numbers of adults found resting in animal shelters near these fields.

65Female An. labranchiae can aggressively attack human hosts, and are described as 'persistent' in their attempt to enter bedrooms during the night. Nonetheless, this species is also described as zoophilic in some of the general literature, but overall, An. labranchiae appears opportunistic in its host choice, readily biting either humans or animals (Table 17). Romi et al. found high percentages (86% and 90.7%) of females engorged with human blood resting inside houses whereas they also found that almost all specimens collected resting in animal shelters had fed on animals.

66Anopheles labranchiae rests inside houses, animal shelters, and, to some degree, in natural shelters, depending on the location of its blood source. D'Alessandro et al. described An. labranchiae as both endo- and exophilic, using whatever shelters are available. Females hibernate in stables/animal shelters and in natural sites such as crevices and tree cavities. Both incomplete (with occasional blood feeding but without ovipositioning) and complete (with fat bodies, without feeding and non-gonoactive) hibernation have been noted for this species.

67As with An. atroparvus, An. labranchiae has been found to be refractory to exotic strains of P. falciparum, with de Zulueta et al. failing to infect An. labranchiae, albeit a small sample, with a Kenyan strain of P. falciparum. However, Toty et al. reported historical evidence of naturally infected An. labranchiae, plus the results of a contemporary study conducted by the Centre de Production et d'Infection d'Anophèles (CEPIA) in Paris where 14% (13/99 specimens) of Corsican An. labranchiae were experimentally infected with the African NF54 laboratory-cultured strain of P. falciparum. This study also detected sporozoites in the salivary glands of three specimens, indicating that An. labranchiae is not only susceptible but also potentially able to transmit at least some strains of African P. falciparum. However, this conclusion must only be considered alongside the knowledge that the NF54 P. falciparum strain is a highly attenuated, long-standing laboratory culture which may no longer reflect its origins (Bangs, unpub. obs.).

68Anopheles messeae is the third member of the Maculipennis Subgroup to be designated as a DVS. It is the most widespread species of the subgroup, with a distribution extending from Ireland across Europe and Asia and into China and Russia. A great deal of work on this species has been conducted in Russia and China. This review is therefore presented with the caveat that there may be details and data reported in the Chinese or Russian literature that are not included here due to access difficulties.

69Di Luca et al. identified a number of genetic polymorphisms within An. messeae and defined five separate haplotypes associated with different geographical areas across its distribution. However, they could not confirm whether these polymorphisms were indicative of altered behaviour at these different locations, although the large range of this species combined with such genetic variability would suggest that some area-specific biological or behavioural adaptations are likely to have occurred.

70The larvae of An. messeae are typically found in shaded, clear, very slow flowing or stagnant, fresh water sites such as lake margins and marshes. Despite only sampling resting females, Adamovic, in Serbia and Montenegro and Adamovic & Paulus, in surveys of Slovenia and Croatia, continually associated the presence of adult An. messeae with stagnant, fresh water oxbow swamps and marshes within alluvial plains or valleys of large river systems and at sites near large lakes. Localities along rivers with saline or alkaline soil did not provide the same association, however they did report the presence of An. messeae at sites near a marshy plain with brackish water. Takken et al. also inferred the presence of An. messeae in more brackish habitats, presenting a photograph in their paper of a drainage ditch labelled as containing brackish water and vegetation which 'supports Anopheles messeae'. However, they also indicated that engineering works in the Netherlands allowed the transition of brackish sites to fresh water, so whether or not they did find this species in brackish water is still unclear. Nonetheless, Takken et al. did identify locations where An. messeae larvae were collected, including sites containing reeds, and those containing floating aquatic weeds and algae, relatively open ditches inside forests and clear water in small lakes within dunes.

71Only one study could be found that sampled An. messeae females inside human habitations, animal shelters and in natural outdoor shelters. All other studies only searched in animals shelters. Where comparisons were made, no An. messeae were found resting outdoors in urban areas (e.g. in vegetation surrounding buildings) but were found indoors such as in entryways, staircases and basements, although not in large numbers. In rural areas, An. messeae dominated the collections made from cattle sheds, with few specimens collected from natural outdoor sites (hollows, ground cavities, amongst vegetation surrounding marshes, ponds and streams).

72Takken et al. argued that in the Netherlands An. messeae has never been considered as a malaria vector despite it being as susceptible to P. vivax infection as An. atroparvus. They stated that its high degree of zoophily and outdoor feeding behaviour makes the likelihood of it being involved in local malaria transmission very remote. They supported their argument reporting that all resting An. messeae females collected in their study had fed on animals; however, all their samples were collected from animal shelters. Bates also mentioned that in Albania, An. messeae (along with other members of the Maculipennis Group) is not considered a malaria vector using the same reasoning: '[An. maculipennis, An. messeae and An. melanoon (as An. subalpinus)] are generally supposed not to be malaria vectors because of their non-anthropophilous [sic] food habits'. Fyodorova et al. found that 40% of the An. messeae females collected in urban areas contained human blood, with the remaining 60% having fed on cats (40%) and chickens (20%). However, in rural areas, no human blood meal was detected. Becker et al. summed up the biting preferences of An. messeae somewhat ambiguously, stating that 'Blood-meals are taken from humans only when the density of An. messeae is very high and there is a shortage of livestock, but they also may attack humans in houses'. No studies were found that examined the feeding cycle of An. messeae.

73Anopheles messeae, like An. atroparvus and An. labranchiae, hibernates as an adult female. However, unlike these other two species, An. messeae chooses hibernation sites in abandoned buildings, in the absence of animals. They enter full diapause, and do not feed during the winter, but instead, gain energy from fat reserves.

74There is some evidence to suggest that, along with An. atroparvus, An. messeae may also be refractory (or essentially refractory) to tropical P. falciparum strains. In their study, testing the susceptibility of Russian anophelines to imported P. falciparum, Daškova & Rasnicyn were unable to infect An. messeae. Indeed, the vector status of An. messeae has come into question, specifically since the discovery of a new species in 2004, formally named An. daciae, in Romania, which has since been recorded from south-western England. Anopheles daciae can only be distinguished from An. messeae using egg morphology or by sequencing the internal transcribed spacer 2 (ITS2) of ribosomal DNA and the cytochrome oxidase 1 (COI) of mitochondrial DNA. It has been suggested that the presence of An. daciae, potentially sympatric across the full range of An. messeae, may be responsible for the high polymorphism previously reported for An. messeae. Combined with an ongoing debate about the capacity of An. messeae to transmit malaria (e.g. it is not considered a vector in northwestern Europe), it is feasible that An. daciae, and not An. messeae, could be involved in malaria transmission, which will only be confirmed with further investigation into the epidemiological importance of each respective species (Harbach, unpub. obs.).

75Anopheles sacharovi is the final member of the Maculipennis Subgroup defined as a DVS and has been the target of a number of focussed, anti-vector campaigns across its range including Israel, Greece and Turkey, yet this species still persists in all areas. Anopheles sacharovi is highly plastic in both adult behaviour and its choice of larval habitats. Zahar states simply: '[An. sacharovi] breeds in all small water collections containing aquatic vegetation'. It makes use of fresh water habitats but is also described as more tolerant of salinity (up to 20%) than any other member of the Maculipennis Subgroup. It can survive in waters up to 38-40°C ( references within), and although it is generally considered to breed in stagnant waters, it can also cope with some, albeit weak, current. Throughout the literature there is general agreement that this species prefers sunlit sites with plenty of emergent and/or floating vegetation. A typical habitat would be an area of swamp or marsh, but larvae are also found at margins of rivers, streams and springs, seepages, wadis, pools and ditches. It is associated with rice cultivation and other irrigated areas, specifically where irrigation channels are poorly constructed causing leakage, creating boggy areas or standing water. Despite its apparent adaptability, An. sacharovi cannot tolerate organic pollutants. Indeed, Saliternik lists the organic pollution of streambed habitats, previously densely populated with An. sacharovi larvae, of greater impact than the wide-scale IRS application of DDT as causal to the near elimination of this species in Israel in the 1960s.

76Anopheles sacharovi females feed opportunistically, despite being generally considered as anthropophilic. Only one study reviewed specifically tested host preference. Demirhan & Kasap, using baited feeding rooms, concluded that in the presence of other, equally available hosts (human, cow, sheep, chicken, horse and donkey), An. sacharovi preferentially fed on donkeys, and had a negative preference for humans. They also analysed the blood meals of engorged females from human habitations, animal shelters and abandoned or ruined buildings and reported the 'feeding preference' of females captured in the human dwellings to be cow, human, sheep, horse and chicken. Other studies reported similar results. Yaghoobi-Ershadi et al. found high numbers of females collected from cow sheds or chicken coops had fed on animals (85.6 - 92.5%), whereas of those collected from bedrooms, only 38.5% had fed on humans, 38.5% on other animals and 23% on both. Boreham & Garrett-Jones reported predominantly human blood in specimens collected from houses, predominantly animal blood (sheep or goat) from animal shelters, and those collected from outdoor pit shelters generally contained blood of mixed animal origin (sheep, goat, horse, dog or cow). Hadjinicolaou & Betzios reported a high percentage of females containing human blood from human habitations, whereas females taken in pit shelters and animal sheds had mostly fed on domesticated animals. They concluded that An. sacharovi still exhibited significant levels of anthropophily despite a high ratio of animals to people (between 9:1 and 7.2:1) in the study area. Boreham & Garrett-Jones suggested that An. sacharovi had increased tendencies towards zoophilic behaviour due to previous DDT spraying campaigns, but was reverting back to anthropophily.

77Anopheles sacharovi, contrary to the accepted night-time biting habits of most anophelines, can 'in deeply shaded situations... attack viciously throughout the day'. However, Djadid et al., indicated that An. sacharovi (plus other members of the Maculipennis Subgroup) generally start biting in the early evening, peaking between 20:00 or 21:00 (refers to Djadid MSc thesis), with Alten et al. noting higher densities of An. sacharovi between 20:00 and 22:00, although they did not specifically examine biting behaviour. Hadjinicolaou & Betzios observed An. sacharovi to bite indoors and outdoors. Biting location is likely to be driven by host behaviour, for example, in the hotter parts of Turkey where both people and animals spend the night outdoors, biting would tend towards exophagy.

78Anopheles sacharovi is principally described as endophilic. Its choice of resting location (and, arguably, for all species) is most likely driven by the need to find the most suitable microclimate for increased survival. Demirhan & Kasap observed An. sacharovi feeding on cows outside, and then entering houses or abandoned shelters to rest. Yaghoobi-Ershadi et al. found An. sacharovi in cow sheds, chicken coops and bedrooms, but were unable to find any females resting outdoors. Boreham & Garrett-Jones searched two artificial pit shelters and found 10 and 42 specimens compared to 377 and 333 from two cattle sheds and 260 in a village house. Abdel-Malek failed to find An. sacharovi resting outdoors, but again, repeatedly found them resting in animal stables and human habitations. However, insecticide residual spraying in many areas has apparently affected endophilic behaviour, summed up by Gokberk: 'Following the last ten years of DDT spraying, An. sacharovi recently began to show a tendency to be less domestic in habits'. Yet, there is evidence that once these IRS programmes ceased, An. sacharovi began to revert to more typical endophilic behavioural patterns.

79As with other European or Middle East DVS that occur in warmer climates, hibernation is incomplete, with intermittent feeding during winter, but without oviposition, often making use of the same localities chosen for resting in the summer months.

80There is some confusion as to the taxonomic status of An. sergentii. It has previously been considered to have two geographical forms: An. sergentii sergentii and An. sergentii macmahoni, but in accordance with the published literature, the Walter Reed Biosystematics Unit (WRBU) online catalogue of the Culicidae and the Mosquito Taxonomic Inventory, An. macmahoni is currently considered a subspecies of An. sergentii. This subspecies has never been found biting humans and is of no known medical importance.

81Anopheles sergentii is known as the 'oasis vector' or the 'desert malaria vector' due to its distribution within oases across the Saharan belt in northern Africa into the Middle East, and its ability to cope with the extreme climate across this region. It may be able to survive in such harsh conditions due to its adaptability. It makes use of a range of larval habitats, including streams, seepages, canals, irrigation channels, springs, rice fields, and most other non-polluted, shallow sites that contain fresh water with a slow current, slight shade and emergent vegetation or algae (Table 14). However, larvae have also been found in moderately brackish habitats, areas of stagnant water, light to moderately polluted locations or in sites in full sunlight. In general, the presence of vegetation or algae seems to be the only characteristic common to all larval habitats of this species.

82Farid described An. sergentii as '...an indiscriminate biter of both humans and animals, both indoors and out.', however, no study could be found that specifically tested host preference. Six studies have reported blood meal analyses of resting mosquitoes, taken from both human and animal shelters (Table 17). Of these, five described An. sergentii as principally or even highly zoophilic, with Kenawy et al. stating the key factor that limits oasis malaria transmission in Egypt is the zoophilic feeding behaviour of An. sergentii. Faraj et al., in Morocco, also described 'a marked preference for zoophily' in this species.

83Kenawy et al. related human biting to local animal stabling practices. They found that in villages where animals were housed in rooms within human habitations, a lower proportion of the An. sergentii females collected resting in the houses contained human blood. In an earlier study, however, the proportion of females with human blood was higher in those taken from houses containing animal rooms (86%), although the absolute numbers (calculated here from percentages reported in the paper) of 18/21 versus 26/54 (48%) from houses with no animal rooms also indicate that An. sergentii were diverted away from human hosts and towards the animals when in close proximity of one another. In this latter study, the animal rooms within the houses had the highest number of resting mosquitoes (209), of which 13% (equivalent to 27 mosquitoes) had fed on humans compared to those in isolated animal sheds (126) with only 7% (equivalent to nine mosquitoes) containing human blood.

84No studies were found that specifically tested biting location, yet Abdoon & Alshahrani reported high numbers biting outdoors and concluded An. sergentii was both exophagic and exophilic after finding few females resting inside houses. However, with no comparable indoor biting data and no sampling of resting mosquitoes from animal shelters, this conclusion may not indicate a true blood feeding preference. Indeed, Saliternik described An. sergentii as feeding and resting both indoors and outdoors, but referred to 'exophilic habits'. Barkai & Saliternik suggested that an 'exophilic strain' of An. sergentii had developed because of indoor spraying with DDT in Israel, finding fewer adults at indoor resting places where they had been common in the past, despite the local abundance of larvae.

85Anopheles sergentii can overwinter as both adult females or larvae, although no details regarding hibernation, blood feeding and oviposition could be found.

86Preliminary data on An. superpictus populations sampled across Iran recently identified three genotypes (designated X, Y and Z) and raised the possibility of An. superpictus as a species complex. These data have yet to be confirmed, but the wide distribution of this species across a number of diverse climatic regions (Mediterranean across to central and southwestern Asia) and the existence of eight junior synonyms, suggests the realistic possibility of An. superpictus being a complex of species and therefore warrants further investigation (Harbach, unpub. obs.).

87In the published literature, An. superpictus larvae are continually associated with gravel or pebble river and stream beds in shallow, slow-flowing clear water in full sunlight. Typical, natural sites are small pools within or next to drying river beds, conditions which are closely related to seasonal fluctuations in precipitation. At such sites, larval abundance increases only in late summer when pools are created as the river levels decline and, once water levels rise with the increasing rain during the onset of winter, these locations again become unsuitable as aquatic habitats.

88Such natural limiting conditions could restrict both the distribution, abundance and period of adult activity of this species, however An. superpictus has easily adapted to human-influenced habitats, making use of irrigation channels and storage tanks and pools formed from their leakage, rice fields, ditches, borrow pits and hoof prints, amongst others. Anopheles superpictus larvae have also been found in brackish water habitats and in stagnant water. Jetten & Takken state that it can occur in polluted sites, although here, no primary data were found to confirm this statement, which is also contradicted by other observations. For example, Berberian, stated that 'A. superpictus is never found in polluted or filthy water...' and the decline of An. superpictus (and An. sergentii) in Israel has been closely associated with sewage pollution of many of the natural streams it previously inhabited. Anopheles superpictus survives at relatively high altitudes, up to 2800 m, replacing An. sacharovi that may dominate at lower altitudes.

89No publication could be found that reports any definitive host preference for An. superpictus, but it is generally given to be a zoophilic species that also readily feeds on humans. Tshinaev reported from Latyshaev (reference unavailable) that, in Uzbekistan 'individuals that who sleep out during the summer on the flat roofs of houses and on towers are not attacked by this mosquito'. Conversely, Ramsdale & Haas stated that An. superpictus in Turkey has a 'marked preference for animals but feeds on man in their absence...', but still described An. superpictus as an 'unusually dangerous mosquito' for those people who spend nights out in the open, away from villages and towns.

90Again, no primary data were found describing biting location. However, An. superpictus appears to be opportunistic in its feeding habits and will enter houses to feed, but is generally regarded as exophagic.

91The BRT model has been applied to contemporary data on the occurrence of 13 DVS in Africa, Europe and the Middle East using the most comprehensive database of DVS occurrence currently available. These maps and the underlying database will be made available in the public domain. We stress that the predictive maps produced will not be perfect representations of the true geographical distributions of these species but nevertheless, they represent a substantive step in improving our knowledge of the range of those DVS studied. One particular issue for any environmental-niche based mapping technique is predictions in areas that, although environmentally suitable, may not contain the vector for other biogeographical reasons. The model predicts, for example, the presence of An. arabiensis throughout Madagascar. Anopheles arabiensis is a species commonly associated with dry, savannah-type habitats and is considered absent, or at least, is rarely encountered in the humid climate of the eastern coast of Madagascar (Manguin, unpub. obs.). Conversely, An. nili has never been recorded in Madagascar (Manguin, unpub. obs.) but, as identified in the predicted distribution, there are areas where conditions are suitable for An. nili to become established if it were to be introduced.

92Biases in collection location, variation in sampling methodologies, limited data for some species or an absence of data over large areas of suspected occurrence all contribute to uncertainty in the final predictions. Yet despite these limitations, the maps represent the first attempt to model DVS distributions across Africa, Europe and the Middle East using a combination of extensive occurrence data combined with contemporary EO distributions. All this information should be triangulated when evaluating the utility of the maps which are best considered as the beginning of an on-going process of understanding, describing and better predicting the range of these DVS. This process may be further complicated by the ever evolving revision of taxonomic status of a number of the African DVS that may lead to further stratification of the occurrence data and revisions of the predictions. This is particularly important where newly identified forms are associated with varying bionomics relevant to their control; the molecular and chromosomal forms of An. gambiae are but one example.

93The behavioural plasticity, large geographic ranges, and changing taxonomic categorisation, in particular with the African DVS, present challenges when summarising the bionomics of individual species. Moreover, conclusions drawn about behavioural characteristics based on biased sampling may mask the true variability in a population and behavioural adaptation to human influences, such as insecticide use or environmental disturbance, can also influence local variation in species bionomics. The bionomics data are again viewed as a significant compendium but with the caveat that expert, local knowledge should always complement the information provided.

94This is the second in a series of three publications describing the distribution and relevant bionomics of the global DVS of malaria. The first publication detailed the DVS of the Americas and the final publication will examine the DVS of the Asian Pacific region (Sinka et al: The dominant Anopheles vectors of human malaria in the Asia Pacific region: occurrence data, distribution maps and bionomic précis, unpublished). Together, these three publications are intended to provide a baseline set of data and maps and summarise the current knowledge of the bionomics of the 41 DVS identified as the primary vectors of P. falciparum and P. vivax malaria.

95The maps and data presented here, and those relating to the DVS of the Americas, and the Asian Pacific region, will be available on the MAP website in accordance with the open access principles of the MAP (please contact authors for details). These data and maps are provided as a dataset to be improved and built upon. Undoubtedly, the process of species distribution mapping will improve, environmental and climatic spatial data will become available at higher resolutions, and more refined understanding of the ecology that limits a given DVS distribution attained. The single most important factor, however, will be more spatially comprehensive occurrence data and this exercise has been additionally valuable in identifying the paucity of information in large areas in Africa, Europe and the Middle East. An increasing willingness to share data between research groups and national malaria control programmes has been instrumental in this initiative and is critical to its sustained future.

96AUC: Area Under the operating characteristic Curve; AVHRR: Advanced Very High Resolution Radiometer; BRT: Boosted Regression Trees; COI: (mitochondrial) Cytochrome Oxidase 1; DEM: Digital Elevation Model; DVS: Dominant Vector Species; EO: Expert Opinion; EVI: Enhanced Vegetation Index; GIS: Geographic Information System; IRS: Insecticide Residual Spraying; ITNs: Insecticide Treated Bednets; ITS2: Internal Transcribed Spacer 2; IVCC: Innovative Vector Control Consortium; LST: Land Surface Temperature; MAP: Malaria Atlas Project; MIR: Middle Infrared Radiation; MODIS: MODerate Resolution Imaging Spectroradiometer; NDVI: Normalized Difference Vegetation Index; PCR: Polymerase Chain Reaction; TAG: Technical Advisory Group; TFA: Temporal Fourier Analysis; WRBU: Walter Reed Biosystematics Unit.

97The authors declare that they have no competing interests.

98SIH conceived the study and managed its design and implementation. MES wrote the first draft of the manuscript and assembled the occurrence data with assistance from CWK and RMO, CWK also digitised and edited all the expert opinion maps. WHT designed and maintained the databases and implemented the map figures. APP implemented the BRT scripts for predictive mapping. PWG processed the environmental and climatic data grids, with assistance from TVB. All TAG members (MJB, SM, MC, CMM and JH) provided data and advice in updating the EO range maps. Experiments were devised by SIH and MES and implemented by MES. All authors participated in the interpretation of results and in the writing and editing of the manuscript. MJB, MC, SM, HCJG, CMM and REH advised on bionomics and nomenclature issues, and provided additional comments and input to the manuscript.