Anopheles larval habitats in Dakar
Several factors were found to be associated with the occurrence and abundance of anopheline larvae that were collected in Dakar from 2007 to 2010. Using multivariate analyses, the probability of the presence of anopheline larvae and the larval density were positively associated with numerous factors that could be classified into the following four categories: climatic factors (rainy season), environmental factors (lower level of urbanization), morphological factors (temporary collections) and biological factors (increased surface vegetation, higher water temperature, higher pH, and the absence of larvivorous fishes). These results were consistent with the known preference of An. gambiae s.l. for breeding in temporary pools [28, 32]. Higher water temperature was not related to the presence of anopheline larvae but was linked to the density of larvae in a collection. The water temperature is perhaps not a key factor in the laying of eggs by parous females; however, it impacts the development of An. gambiae immature stages. Low-floating vegetation was previously found to be a determinant of the presence of anopheline larvae [12, 33]; however, in the present study, the level of floating vegetation was positively associated with the presence of larvae. This difference could be explained by the difficulties of measuring the surface of vegetation that covers a water collection. Also, the surface vegetation can be a proxy for the presence of food, which can favour the presence of anopheline larvae. The pH was significantly associated with the occurrence or the abundance of larvae, and this factor could also be considered to be an indicator of the presence of food for larvae. The existence of a muddy bottom in and around the water collection was favourable for the presence of larvae, and this observation was consistent with the possibility for An. gambiae to lay their eggs on soil around puddle larval habitats [34, 35]. However, a muddy bottom could also be a proxy for an increased persistence of water, which is different from asphalt or sand. As was previously described, the presence of predator fishes was associated with a lower probability of larvae or lower larval density. In 1998 in Dakar, Awono-Ambéné et al  confirmed that predation in the "céanes" was probably mainly due to fishes (Gambusia and Tilapia) that were introduced in Dakar in the 1930s for larval control. Many works have been conducted on the impact of urban agriculture on malaria transmission [11–15, 37, 38], and most of these studies have shown that this practice was associated with a higher level of malaria transmission in the surroundings areas. Specifically, urban agricultural activities can provide breeding and resting sites for malaria vectors. The results of the present study tend to show that market-gardens provided resting sites for adult Anopheles, rather than increased the number of larval habitats, as was previously demonstrated in Ghana . Indeed, the water collections that were located in market-gardens less frequently harboured anopheline larvae, and when anopheline larvae were present, the densities were lower than in surroundings areas (as determined by univariate analysis). This occurrence was probably due to the presence of larvivorous fishes in traditional wells and to the frequent perturbations of water due to watering; however, one cannot exclude that this effect occurred because of the use of pesticides by urban farmers [39, 40].
Anopheles larvae in public and private areas
The present study has found that few larval habitats were located inside private sectors, in comparison to public areas (i.e., freely accessible areas). Regardless of the type of housing or socio-economic level, the population generally did not allow Anopheles breeding sites to appear and persist in their neighbourhoods. As a consequence, the efficacy of any larval control measures would not suffer by being focused on freely accessible public zones, and excluding private areas from this treatment could simplify the control procedures.
Spatial heterogeneity in malaria transmission in Dakar
The present results provide new evidence for malaria transmission in downtown Dakar and its nearby suburbs. The spatial heterogeneity of HBR was very marked, and it ranged from 0.1 to nearly 250 bites per person per night during the rainy season. According to the analysis undertaken in the studied areas, most of the adult Anopheles were caught between July and December, and a marked peak occurred in September-October. From 2007 to 2010, An. gambiae s.l. that had been infected with P. falciparum were caught in 22 of the 45 studied zones. Combined with the investigations that were conducted in 2005 and 2006 , it was determined that infected An. gambiae s.l have been caught in 24 of the 45 studied zones. Among the 14 areas where infected An. gambiae s.l. have been caught, no statistical difference has been highlighted between the CSP indexes for two consecutive years, indicating that no major change could have been detected in the infection levels in Dakar. Even if the CSP indexes had been calculated with the figures from the rainy season in 2007 and the annual figures in 2008-2009 and 2009-2010, it has been possible to compare data because the present study showed that the An. gambiae s.l. bites were concentrated in September and October, with a minimal HBR thereafter (constantly less than 5 bites per person per night from November to July). Consistency between the mosquito collections (and the HBR evaluation) was also ensured during the study period because the protocol for sampling remained unchanged. In addition, no major environmental changes have been recorded in the studied areas because of the low evolution rate in previously urbanised areas, which allows for the comparison of adult mosquito figures between years.
The ELISA method for the detection of CSP antigen in malaria vectors has been proven to detect more positive mosquitoes than dissection of salivary glands. However, the ELISA method is subject to less inter-operator variations and is less time-consuming than salivary gland dissection. For these reasons, ELISA is often considered to be the most reliable method for estimating the proportion of Plasmodium-infected malaria vectors and for analysing malaria transmission.
Anopheles species involved in malaria transmission in Dakar
The following five anopheline species were collected from the 45 studied areas of Dakar: three species from the An. gambiae s.l. complex, Anopheles pharoensis and Anopheles ziemanni. Anopheles ziemanni had been previously caught on human bait at low levels in Senegal , as was the case in the present study, probably because of the low affinity of this species to human blood. Anopheles pharoensis is considered to be a secondary vector for malaria and has been found in the present study at low density. The presence of An. gambiae s.s. M form was reported in 2008-2009 (and in 2005-2006 in ) at extremely low levels. Therefore, An. gambiae s.s., An. ziemanni and An. pharoensis could not be significantly involved in malaria transmission in Dakar.
From 2007 to 2010 in the 45 studied zones, 54 An. gambiae s.l. were infected by P. falciparum: 52 An. arabiensis and two An. melas. Both infected An. melas were caught in Golf, near the large marshy area, in 2010, and this was the first time that this species was found to be infected with P. falciparum in Dakar. From 2007 to 2010, An. melas has been reported in 18 zones, where it accounted for 1% to 40.9% of the An. gambiae complex, according to the area and the year. Because An. melas is not as good a vector as An. arabiensis , it could be involved in malaria transmission only in areas in Dakar where it is very abundant. None of the breeding sites were exclusive for An. melas during the entire duration of the survey, and An. arabiensis was always found at these sites on at least one date, probably because of variations in the salinity levels of the breeding habitats. The co-occurrence of both of these species indicates that An. arabiensis larvae may also have a tolerance for low salinity.
Biting behaviours of adult Anopheles
In 2007, 2008-2009 and 2009-2010, 21%, 30% and 26%, respectively, of An. gambiae s.l. were caught indoors (at one indoor versus two outdoor catching points), but it was not possible to define a specific behaviour according to the season or to the studied area. Because logistical constraints limited the number of indoor catching points to one per site and the characteristics of the rooms could be different between areas (e.g., public offices or private properties of different sizes and types), the experimental design may have not allowed for the results to be representative of the adult An. gambiae s.l. behaviour in each of the studied areas. It can only be concluded that malaria transmission can occur in Dakar both outdoors and indoors. No difference was recorded in the hours of activity of An. gambiae s.l. between the different studied sites, and most of An. gambiae s.l. bites (> 75%) were recorded to take place after midnight. This finding highlighted the interest of using insecticide-impregnated bed nets to control malaria transmission in Dakar. Nevertheless, some bites were received during the first hours of the night when people were generally not asleep and not protected by mosquito nets. The use of other mosquito-control devices, like repellents or mosquito coils, could be proposed to increase protection for people from mosquitoes during the first hours of the night in Dakar.
Temporal heterogeneity of malaria transmission in Dakar
The abundance of An. gambiae s.l. during the year was clearly dependent on rainfall because the peak of the An. gambiae s.l. biting rate followed the peak of rainfall, with a two-week lag, regardless of the year. Afterward, the densities quickly decreased after the rains ended, and only isolated specimens were caught until the beginning of the new rainy season in July. Nevertheless, in one area (Golf), some infected mosquitoes were caught during the dry season, indicating that a permanent malaria transmission is possible in this area of Dakar, which is located near the large marshy area. When comparing the human biting rates in the common areas between 2007, a year with low rainfall, and 2008, a year with heavy rainfall, the peaks of Anopheles bites were, on average, six-fold higher in 2008 than in 2007. This was consistent with the fact that rainfall is closely related to the presence of surface water, larval productivity and adult densities, the latter two being highly correlated. Indeed, several studies have highlighted the low dispersion of Anopheles in urban settings due to the high density of houses and the proximity of readily available hosts for blood meals; therefore, most An. gambiae s.l. that are caught on human bait probably hatched from eggs that have been laid in a breeding site in the same or neighbouring area [15–19].
Toward malaria risk mapping
Most of the factors that are associated with the presence and density of larvae in water collections are measurable or detectable, either directly or indirectly, by remote sensing (RS) and include the following: the soil temperature, vegetation and trees that are able to provide shade, the surface vegetation, the moisture level of the ground, the density of urbanization, and the persistence of water collections during the dry season. Consequently, RS and geographic information systems (GIS) can provide useful information for mapping Anopheles breeding sites, as has been shown in numerous studies . Furthermore, the close relationship between larval density and adult HBR suggests that maps of Anopheles breeding sites could provide baseline data for mapping Anopheles adult densities.