Background
Global statistics indicate that malaria morbidity and mortality have declined mostly as a result of scaling up vector control interventions [
1,
2], but that the gains are stagnating in some countries [
3]. This decline has also been observed in Tanzania with more than 50% reduction of malaria burden recorded since the year 2000 [
4]. The 2016–2017 Tanzania malaria indicator survey demonstrated reduction of prevalence in children under five, from 18.0% in 2008 to 7.3% in 2017 [
5‐
7] in the mainland. These successes may be attributed to scaling up of vector control tools, such as long-lasting insecticide-treated bed nets [
8] and indoor residual sprays [
9], as well as socio-economic developments and urbanization [
10] and treatment with artemisinin-based combination therapy (ACT) [
11].
In a recent study, it was demonstrated that just one malaria vector species,
Anopheles funestus, contributes to more than 85% of malaria transmission in south-eastern Tanzania, despite occurring at far lower densities than the other major vector,
An. arabiensis [
12]. Unfortunately, the
An. funestus populations are resistant to pyrethroid insecticides commonly-used on bed nets [
12,
13], survive longer than
An. arabiensis, and feed almost exclusively on humans [
12,
14]. Other studies in different African countries have also documented
An. funestus resistance to pyrethroids: Malawi [
15,
16], Mozambique [
17,
18], South Africa [
19], Zambia [
20,
21], Zimbabwe [
21], Cameroon [
22,
23] and Senegal [
24], a situation that compromises effectiveness of current vector control options [
25], and perpetuates residual transmission even in communities where bed net coverage is more than 90% [
26]. Given its dominance in Tanzania, it has been suggested that interventions that effectively target
An. funestus could have a high impact on residual transmission here [
12]. One potentially effective approach involves suppressing mosquito populations by identifying and directly targeting
Anopheles swarms with highly effective insecticides [
27], before the mosquitoes enter houses.
Swarming behaviours have previously been intensively studied, mostly in Western and Central Africa [
28‐
30]. However in East and southern Africa, there have only been a few studies in Zambia and Mozambique [
20,
31] in addition to an old set of observations in northern Tanzania in the 1980s [
32].
Anopheles swarms are naturally difficult to find since they occur at dusk when visibility is poor and last only for a few minutes. In fact, the rarity of
Anopheles swarms in East Africa has led some to hypothesise that they mate primarily indoors, such as previously observed in experiments by Dao et al. [
33]. Nevertheless, a recent study by the Ifakara Health Institute, which relied primarily on community volunteers rather than experts, demonstrated natural occurrence of swarms of
An. arabiensis in villages across south-eastern Tanzania [
34], more than three decades after the last records of this species swarming anywhere in the region [
32]. In addition to providing detailed characterization of over 200
An. arabiensis swarms, the study also demonstrated that trained volunteers were able to identify and locate mosquito swarms in their villages [
34]. Follow-up studies have since demonstrated potential for localized control by targeting the swarms using aerosol spraying (Kaindoa et al. unpublished data).
An interesting revelation in that last study on
An. arabiensis swarms was a single incidence where 13
An. funestus males were caught in a sweep net [
34], providing earliest indications that this species too formed swarms in the valley, but that these swarms were certainly more elusive than those of
An. arabiensis. The aim of this current study was, therefore, to identify and characterize swarms of
An. funestus, which is now the dominant malaria vector species in rural south-eastern Tanzania.
Discussion
Improved understanding of malaria vector ecology and behaviour is crucial for achieving the ambition of malaria elimination through vector control tools [
41]. The swarming and mating behaviours in particular are components of mosquito behaviour that has been neglected for a long time [
30,
31,
34]. Given that
An. funestus contributes to more than 80% of the ongoing malaria transmission in the area, it is important to identify and characterize swarms of
An. funestus, so as to explore complimentary tools, possibly targeting swarming behaviour.
Furthermore, it is important to understand mating behaviour for implementation of certain control methods, e.g. use of sterile insect techniques and release of genetically modified males. This is the first verified report of
An. funestus swarms in Tanzania, though the earlier study confirmed natural occurrence of
An. arabiensis swarms in the same area [
34]. Together, the two reports confirm widespread occurrence of swarms by the main malaria vectors, even though they are generally elusive and require dedicated teams with local knowledge to map.
An attempt to characterize the swarm sites did not yield any obviously distinctive physical markers of the swarm stations. Instead, most of the swarms were observed to occur above bare ground, sometime on the front lawns of human houses (Fig.
4). This is very similar to observations by Charlwood et al. who suggested that
An. funestus mosquitoes in Mozambique formed swarms close to the houses used for resting and that the swarming sites could be used as indicators of houses to be targeted for vector control interventions [
31].
Proximity to houses is therefore an essential condition for
An. funestus mating swarms, and the possibility that some degree of mating happens indoors cannot be excluded. Relative to previous observations in the same area [
34], this current study has demonstrated that
An. funestus swarms differ from those of
An. arabiensis in terms of height, swarm size and location. For example, while
An. arabiensis males were observed swarming at mean heights of 2.5 m, swarms of
An. funestus occurred at much lower heights averaging just 1.7 m above ground, with several swarms lower than 1 m. In terms of location and markers,
An. funestus preferred to swarm very close to human houses, unlike
An. arabiensis swarms, which were mostly at the edge of the villages [
34], possibly due to the greater anthropophily of the former than the latter species. The
An. funestus swarms were small and generally consisted of less than 15 mosquitoes, as collected by sweep nets, while
An. arabiensis swarm sizes ranged from approximately 10 to 60 mosquitoes [
34]. It is possible that these differences in vector densities may be associated with season (this
An. funestus study was conducted between June and July 2018 while the
An. arabiensis study was conducted between August 2016 and June 2017). Nonetheless, the natural differences in population sizes as previously observed in the valley [
12].
While the work with community volunteers certainly increased the ability of research team to identify the
An. funestus swarms, visual estimates of the swarm sizes by the volunteers did not strongly correlate with the sweep net estimates (R = 0.4518). Though the volunteers were able to locate
An. funestus swarms, either they could not accurately estimate the number of mosquitoes in the swarm, or the flying males were able to avoid the sweep nets. In the previous study however, there was a strong correlation between sweep nets estimates and visual estimates of the number of
An. arabiensis in the swarm, indicating that visual estimates could possibly be relied upon to estimate swarm sizes [
34]. Given that this current study was not done across multiple seasons, the discordance between estimates may indicate density dependence of such relationships, or that the correlations are non-linear (Fig.
3). Besides, one limitation of the approach is that the approximation of the peak swarming time and the collections using sweep nets may have been imprecise, even though the standardization allowed comparison across all the swarming events.
The molecular analysis confirmed that 100% of all amplified samples of
An. funestus s.l. mosquitoes were
An. funestus sensu stricto (s.s.). Although
Anopheles rivulorum and
Anopheles leesoni have also been recorded from this area [
12], the present study did not find any other sibling species in the swarms apart from
An. funestus s.s. However, in a recent study in Zambia, a mixed swarm of
An. funestus and
An. leesoni was found [
20]. Though rare,
An. arabiensis and
An. funestus mosquitoes may also swarm either in very close proximity or together. Indeed, in the previous study of
An. arabiensis swarms, one instance of 13 male
An. funestus mosquitoes was observed in a sweep net targeting the former species [
34]. Also, in this current work, four instances of 2, 2, 3 and 7
An. arabiensis males were caught in a sweep net targeting
An. funestus. Previously, mixed swarms of
Anopheles coluzzii,
An. gambiae and
An. funestus have also been reported from other areas in Africa [
20,
42].
Genitalia rotation is a physiological change that occurs when male mosquitoes become sexually mature [
20]. Dahan and Koekemoer indicated that these are visible a few hours after emergence, but the rotation rate can increase with the increase in temperature [
38]. During this process, the genitalia turn clockwise or ant-clockwise until sexual maturity is reached at 180 degrees full rotation. This current study showed that 100% of all sampled males had complete genitalia rotation, suggesting that only sexually mature males participate in the swarming activity.
It was also observed that there was no special selection based on size of mosquitoes entering the swarms. The wing lengths of
An. funestus males ranged between 2.0 and 2.8 mm, with no statistically significant differences observed in mean sizes between the mosquitoes caught in the swarms and those caught resting indoors. The results are comparable with those of Charlwood et al. [
31] on
An. funestus in Mozambique.
The elusive nature of
Anopheles swarms in East Africa is confirmed the paucity or complete lack of reports on such swarms by previous entomologists working in the region. However, this study adopted the approach of working with trained community members to search for swarms as previously described by Kaindoa et al. [
34], and also in Burkina Faso [
27]. Similar approaches have also been used by community members to accurately identify places with low, medium and high mosquito densities [
43]. This has implications for vector control strategies using community participation in targeting mosquito swarms. Indeed, there are several examples where community participation in vector control programs have been successfully relied upon for disease control [
44‐
46]. Even though aerosol spraying could be used to target swarms, additional surveys are needed given that
An. funestus swarms occur very close to human houses and are generally smaller than
An. arabiensis swarms. Moreover, additional safety precautions would be required to protect humans. Alternatively, improved technologies such as small robotic drones could potentially be used to target the identified swarms of
An. funestus mosquitoes, and apply small but targeted insecticide doses.
The challenge of identifying An. funestus swarms in the study area was associated with the unexpected low height of the swarms, given that this study had used previous knowledge of An. arabiensis swarms when searching for An. funestus swarms. Anopheles funestus swarms occurred in close proximity to the houses, places that could not be predicted or associated with mosquito swarming. Additional research and exploration of technologies such as the use of unmanned aerial vehicles fitted with high-resolution infrared cameras could help to locate swarms in areas that are inaccessible by humans. Moreover, though the characterization here was restricted to one season, future studies should consider assessments across multiple seasons to assess whether climatic factors may have an influence on the characteristics of these An. funestus swarms. There is also a need to develop methods for prediction and estimation of An. funestus swarms, which could help to improve the control of malaria in rural areas.
Authors’ contributions
EWK and FOO conceived the study and developed study protocols. EWK, HSN, ALJ, EH, SA, JK and MT supervised data collection. EWK, HSN, AJL, SA conducted data analysis, EWK, HSN, AJL, EH, SA, AM, RM, GM, HB, MC, and FOO wrote the manuscript. All authors read and approved the final manuscript.