Larval source management (LSM), a strategy which includes larviciding and source reduction (environmental manipulation, modification and elimination of aquatic habitats) for mosquito larval control, has long been used as a measure for malaria control in many parts of the tropics [1–7]. More recently larval control has been shown to be highly effective at reducing malaria transmission in Eritrea [8, 9], Kenya [10, 11] and Tanzania [12, 13]. These studies were all carried out in sites with well-defined breeding sites that were not too extensive. However, it is not known whether this approach will be effective in areas where breeding sites cover large areas, as in the floodplains of large rivers or lakes. The work presented here is part of a series of studies recently undertaken in the The Gambia to determine whether larviciding with microbials [14] can reduce malaria transmission in an area of extensive seasonal flooding and to design rational application strategies for operational implementation.

In The Gambia, the landscape is dominated by the river and its floodplains dividing the small country into the river's north and south bank. Seasonal flooding creates large areas of water for extended periods of time and provides potential breeding sites for mosquitoes including the primary and secondary malaria vectors Anopheles gambiae, Anopheles arabiensis, Anopheles melas, Anopheles funestus, Anopheles coustani and Anopheles pharaoensis [15]. In The Gambia, malaria transmission is highly seasonal and peaks at the end of the rainy season which typically takes place from June to October [16, 17]. In the middle reaches of the river in The Gambia (approximately 150 km from its mouths) there is considerable local variation in malaria intensity between villages. Malaria prevalence in children ranges between 5–90% [16, 17] and the average entomological inoculation rate (EIR) between 0–150 infective bites per person per year [18].

The national strategy for malaria control in the country includes mosquito larval control [19], yet there has been no detailed evaluation of this methodology until now. The common malaria vectors in The Gambia are highly susceptible to commercial formulations of microbial larvicides in laboratory and field trials but the larvicides failed to exhibit any residual effect, which means that weekly applications are necessary to prevent adult mosquito emergence [14].

Mapping of all aquatic habitats in a study area of 400 km² was implemented under operational conditions [20] as a pre-requisite for successful anti-larval interventions [12, 21] This was done to guide the larval control programme and determine whether specific habitat characteristics were associated with the presence of anopheline larvae [20]. It was hoped that any such characteristics could be used to guide interventions to target LSM at specific sites to reduce the logistics and costs necessary to implement larviciding at weekly intervals. The mapping revealed [20] a large number of shallow water bodies during the rainy season, primarily in the floodplains. The risk of finding anophelines increased in habitats located within the first one km stretch of the floodplains, from the landward edge towards the river. These were large in size and located in areas where grassy vegetation (including rice and sedge) dominated the land cover. Unfortunately, over 80% of all habitats shared similar features and these could not be used as criteria to help target anti-larval interventions. Thus, the type of water body and the habitat in which they were located could not be used for identifying sites for targeted LSM. Nevertheless, at any time of the year, only a small percentage of aquatic habitats actually contained mosquito larvae [20] suggesting that other criteria than those surveyed under operational conditions may be responsible for habitat colonization and productivity, such as microclimate, water quality and interactions with other aquatic organisms.

This experimental study was carried out to determine whether a detailed ecological and physical characterization of anopheline habitats could explain differences not only in vector colonization but also in larval density and, more importantly, densities of emerging adults in order to determine whether larval control can be targeted at particular sites [22].

The identification of aquatic habitats preferably colonized by malaria vectors, and those most productive, remains a challenge since our results show that apparently similar habitats would not always contain anopheline larvae. Only two factors increased the probability of finding anopheline larvae in a habitat; namely the presence of tufts of grass or other short emergent vegetation and the presence of algae. Both factors have been identified to be of importance in a number of other studies throughout Africa [30–34]. The presence of algae in anopheline production remains controversial. Whilst algae can provide a viable food source for Anopheles larvae [32], algae are frequently associated with older habitats which may be less suitable for Anopheles development [30].

The density of larvae was dependent on a number of physical characteristics. Average larval densities decreased with increasing coverage of the water surface by tall vegetation, an observation that has also been made in various other environments throughout Africa [7, 33, 35] and was historically utilised for mosquito larval source management [7, 36–39]. Larval density also decreased when the water was tidal which was the case for the majority of larger (>100 m perimeter) and deeper (>50 cm) habitats. These habitats were also characterized by a high water conductivity (>2000 μS/cm), which specifically reduced the larval density of the primary malaria vectors An. gambiae s.s. and An. arabiensis. These factors probably explain the observation made earlier in a longitudinal survey implemented under operational conditions [20] that larvae were more likely to be found in the first 1 km stretch from the edge of the floodplain than closer to the river.

Anopheles arabiensis was the most prevalent species of the An. gambiae complex in the study area, which is in agreement with previous findings from dry and humid savannah areas [40, 41] but this is not reflected in the indoor collections from houses near the larval habitats, where An. gambiae s.s. dominates [42]. This might be due to the fact that An. arabiensis is a more exophilic mosquito and feeds to a large proportion on cattle which can be found closer to the habitats [43, 44]. Nevertheless, the dominant Anopheles species in this survey was An. coustani s.l., a mosquito that rests and bites outside [45]. Anopheles coustani and An. pharaoensis are considered secondary vectors of malaria in The Gambia but no published information is available about their role in malaria transmission in the country. Notably, An. coustani and An. pharaoensis represented a large proportion of the sampled anophelines; that is 44% of the identified larvae and 70% of adults. Recent studies from Senegal have shown that environmental changes can lead to changes in the vectorial system [41, 46] and these may be further impacted if vector control strategies, which focus on indoor host-seeking females lead to a shift in species composition. In such events outdoor biting and resting mosquitoes may increase in importance as vectors of malaria [47].

Most primary and secondary adult malaria vectors emerged in August and September. The average An. gambiae production was significantly higher in confined pools than in all other habitats except in rice fields at the edge of the floodplain close to the upland, which share a number of characteristics with pools since they are more defined than other habitats, have high turbidity and more frequently contain algae. Anopheles coustani emerged in higher densities than An. gambiae in all habitat types surveyed. The overall productivity per m² of aquatic habitat in The Gambia was very low for vectors and other invertebrates alike. In relatively small and confined An. gambiae habitats in Kenya for example 18–200 larvae were sampled per area sampler (notably the area sampler was much smaller; 78.5 cm²), and 1–30 pupae collected per m² depending on habitat type and season [48]. In the western Kenyan highlands, a one year survey with emergence traps [49] estimated the average productivity of aquatic habitats to be 1.82 An. gambiae per m² per week; 14-times higher than the findings from The Gambia.

Confined habitats were most likely to have high densities of larvae and lead to the greatest number of adults emerging. They were typically characterized by freshwater with low conductivity, high turbidity, presence of algae and the lack of disturbance by tidal water movements. Vectors and other invertebrates alike thrived in this environment. This is probably a result of the protection that these habitats provide from changing environmental impact due to tidal water movement and the high nutrients present in these habitats. Freshwater streams ideally have a conductivity between 150–500 μS/cm to support a diverse aquatic life [50]. However, extremely high water conductivity due to high salinity and other dissolved ions, as measured in most habitats in this region of The Gambia, has a negative impact on primary production [51]. Run-off, plants, phyto- and zooplankton and other microorganisms are responsible for the provision of food for all primary consumers (e.g. chironomid and mosquito larvae), which serve as food sources for secondary consumers (e.g. predatory dragon- and damselfly larvae, bugs, fish). The vast floodwater area of this region of the Gambia River is poor in nutrients presenting little food for both, primary and secondary consumers. It is, therefore, not surprising that mosquito larvae were found associated with other invertebrates which are primarily predatory organisms. This extends also to the presence of fish in some habitats. In rice fields for example, the presence of small fish was positively associated with the density of vectors and other invertebrates (OR 2.883, 95% CI 1.141–7.284, p = 0.025). Invertebrate predators and small fish are natural enemies of mosquito larvae [52–54] and ovipositing mosquitoes are known to avoid habitats colonised by predators [55, 56]. These apparently contradictory findings may indicate that greater reproductive success can be achieved by ovipositing females laying their eggs in water bodies with predators, where the chance of producing offspring is greater than that found in less salubrious water bodies, where there are no predators. Here water quality is perhaps more important than the presence of predators.

Two aspects affect larval distribution in aquatic habitats; the oviposition choice of a gravid female and the survival of larvae in the aquatic environment. Early instars of anopheline larvae were found in 88% of all habitats where larvae were present (cases), whilst late instars occurred only in 59% of sites. This result may indicate that oviposition occurs in a larger number of habitats perceived suitable by the ovipositing female but that the survival of larvae as expressed by larval density depends on factors associated with habitat size, stability and conductivity of water bodies. Another explanation for the wider distribution of early instars could be that the early instar larvae are more readily dispersed by the movement of tidal water than late instars. Binary analyses revealed that the presence of short vegetation and algae at the sampling point was associated with the presence of early, but not late instars, indicating that these factors, or their associated covariates like nutrients, bacteria and volatiles released from the water [57], are important for the adult female seeking oviposition sites, but that additional factors impact on larval survival.

Rice is the staple food for people in The Gambia and in this part of the country rice fields represent the majority of habitats. After the seasonal rains start in June, women plough their fields in the floodplains of the Gambia River and prepare raised beds for growing rice. Two sets of young plants are transplanted to the fields. The first are transplanted early on in the rainy season in paddies close to the landward edge of the floodplain in August. These are the first areas to flood after the soil becomes saturated following several heavy downpours. However, by October, most of them are drying out and the second set of young rice plants are transplanted to fields close to the river from late September to early October. Here the water is largely flowing in from the river. This explains the seasonality in adult emergence observed. Fields closer to the edge of the floodplains produced An. gambiae adults from the end of August until the end of September, whereas fields closer to the river produced adults at the end of the rainy season in October/November. The overall adult productivity over the season was similar in both fields far away and close to the river, which contradicts findings from a far more intensive but smaller study closer to Farafenni which found nearly all adults emerged from rice fields located within 100 m from the landward edge, away from the river (Jarju, L. personal communication). These contradictory findings suggest that within the smaller study area stronger tides (since it is nearer the river's mouth) may have reduced mosquito production close to the river. Distance from the edge of the floodplain was not a limiting factor either, demonstrating that adults are produced from within the flooded areas and not just on the landward fringe.

Over the entire sampling period, only 33 mosquito pupae were collected with area samplers. This confirms earlier observations [14] that pupae are rare, widely distributed and difficult to sample in the extensive habitats of The Gambia. The use of area samplers did not improve the sampling success in comparison to dipping and therefore cannot be used as a reliable tool to estimate emergence. Generally, the difficulty in sampling vectors in the highly mobile environment in the floodplain of The Gambia, especially closer to the river, needs to be emphasized. The productivity in tidal habitats might have been underestimated in contrast to defined pools using emergence traps since larvae travel with water and might have not emerged under the trap. It is simply far easier to sample mosquitoes in small, defined habitats with higher density per surface area than in large and thinly populated habitats with constant water movement.

In an extensive habitat survey of an area of 400 km², irrigated rice fields presented the vast majority of aquatic habitats with a total surface area of 4,300 ha (43 km²), the landward edge of the floodwater accounted for approximately 1000 ha (10 km²) of aquatic habitats whilst in total all pools covered a surface area of 2 ha (0.02 km²) only, representing only 0.03% of the potential mosquito breeding sites [20]. Therefore, highly productive breeding sites, like confined pools, are small in surface area in this part of The Gambia, whereas those with low productivity are large and widely distributed, and can be found throughout the rainy season. The vast majority of potential breeding sites are floodwater areas including extensive swamp rice cultivations over the entire width of the floodplain. All habitat types are equally colonized and even though the majority of sites have low larval density resulting in a low emergence rate, these tidal floodwater areas including rice fields present the majority of surface water in The Gambia. Even though risk factors could be identified for the presence and density of larvae and productivity of habitats, they indicate that anti-larval interventions in this part of The Gambia cannot be targeted in space or time during the rainy season.

Persons shown in the photographs have consented to publication.