Background
Sampling adult female
Anopheles mosquitoes is crucial for studies of bionomics, to estimate population parameters, quantify malaria and filarial parasite transmission, and evaluate vector-targeted disease interventions [
1‐
3]. Three common sampling methods are (i) the human landing catch (HLC), (ii) baited or unbaited light trap (LT), and (iii) resting collection (RC). These can be conducted in various locations such as within human or animal domiciles or amongst vegetation. The HLC method involves human volunteers luring and capturing host-seeking mosquitoes as they land on exposed legs [
3]. The method provides direct estimates of the human-biting rate (HBR) and infectious biting rate (IBR), and provides a means for characterizing such important bionomic properties as nocturnal periodicity of the biting cycle [
2,
3]. The LT method involves the use of battery-powered suction devices fitted with light bulbs and/or artificial host odors to attract and trap host-seeking adult mosquitoes. The method provides estimates of relative mosquito density, indirect estimates of IBR, and can be used to assess species diversity, community composition, relative abundance, and distribution [
1,
4‐
9]. The RC method involves search and capture of endophilic mosquitoes settled inside human houses or animal sheds, and exophilic mosquitoes resting in the surrounding vegetation or in intentionally placed resting shelters [
1,
3,
10]. It may involve spraying insecticides that rapidly knock down mosquitoes in indoor spaces [
10]. Such collections are used to study mosquito resting habits, recover blood-fed mosquitoes for analysis of host selection, to analyse distribution in space and time, and to evaluate the effect of residual insecticide treatments on endophilic vectors [
3,
11].
These mosquito sampling methods have been applied in Papua New Guinea (PNG)—a country where malaria is endemic with all four solely-human malaria species present and where the IBR can exceed 1000 infectious bites per person per year in some locations [
12,
13]. Over 20 different species of
Anopheles are found in PNG [
5,
14,
15], of which 11 have been incriminated as vectors of human malaria [
14]. Seven of these species, namely
Anopheles farauti sensu stricto (
s.s.),
Anopheles hinesorum,
Anopheles farauti no. 4,
Anopheles punctulatus s.s.,
Anopheles koliensis,
Anopheles longirostris and
Anopheles bancroftii, are vectors of malaria in lowland areas of PNG, including Madang province, where malaria is highly endemic [
13,
16‐
18]. The first five are major vectors whereas the last two play a minor role in the transmission of malaria in PNG [
13,
14,
16,
17,
19‐
21]. Various aspects of these vector species including HBR, IBR, nocturnal biting cycle, dispersal range, and survival rate were studied using one or combinations of these methods [
13,
16,
18,
21‐
23]. Studies of host selection relied on indoor and outdoor RC to recover blood-fed mosquitoes [
17,
23,
24], but this approach has serious limitations in the PNG setting. All of the species tend to be exophagic [
25], thus few individuals (mostly human-fed ones) are found resting indoors, resulting in insufficient and biased samples. The wide dispersal range of some of these species along with thick tropical vegetation makes the outdoor resting search for mosquitoes in peri-domestic environments a laborious task, often resulting in very few mosquitoes that also do not adequately represent the population [
23].
The barrier screen sampling (BSS) method was developed as an alternative to the methods discussed above [
26‐
28]. It involves the use of agricultural shade cloth positioned vertically around villages and imposing a physical barrier suitable for temporary landing and resting of
Anopheles mosquitoes as they commute into and out of the villages. Mosquito collectors visit the barrier screens at specific intervals throughout the night to capture the resting mosquitoes. Unlike the indoor resting collection, both anthropophilic and zoophilic mosquitoes may be intercepted as they rest on the barrier screen. By sampling the mosquitoes as they commute into and out of a village throughout the night, the BSS method overcomes the laborious task associated with outdoor resting collection. The flexibility in screen placement allows sampling in various locations which reduces bias associated with particular sampling locations. The effectiveness of this method to produce a sample of blood-fed mosquitoes for estimating host selection tendencies of several species of
Anopheles vectors of malaria in PNG has been reported [
28]. However, its application for studies of other behavioural or ecological aspects of
Anopheles populations in PNG has not been reported. In this study, the BSS method was used to analyse species abundance, composition and nocturnal movement pattern of
Anopheles species in the coastal and inland lowland malaria-endemic areas of PNG.
Discussion
The effectiveness of the BSS method to produce an adequate and unbiased sample of blood-fed mosquitoes for estimating host selection tendencies of mosquito vectors, and to study timing of host-seeking and infer flight behaviour, has been reported for several
Anopheles species of the southwest Pacific, including PNG [
26‐
28]. The
Anopheles fauna encountered in our study villages using this method of sampling was limited to seven species, but species composition and dominance varied considerably, with certain species dominant in some villages (e.g.,
An. farauti no. 4 in Kokofine,
An. farauti s.s. in Mirap) while completely absent (
An. farauti no. 4) or present but relatively less dominant or common (
An. farauti s.s. in Dimer and Wasab) in other villages. These findings are consistent with, and extend those, of two other studies using other sampling methods [
13,
32]. In particular, the dominance of
An. farauti s.s. in the coastal villages was reversed by greater relative abundance of
An.
punctulatus s.s. in the nearby inland villages (Fig.
2a).
Anopheles longirostris was present across all sites at least in some years and typically a less common species, but was dominant in samples from Wasab for two of the 3 years (Fig.
2b).
The observed diversity of the species combined with variation in their ecological attributes [
23,
24,
28,
40] can potentially attenuate single-intervention malaria vector control programmes such as long-lasting insecticide-treated bed nets (LLINs) in PNG. For example, a previous study showed that
An. koliensis is highly anthropophilic whether or not LLINs were distributed in communities, whereas
An. longirostris,
An. punctulatus s.s.,
An. farauti s.s. and
An. farauti no. 4, were more plastic in their host selection tendencies and diverted more feedings to pig and dog hosts when LLINs were in use [
28]. Indeed, other studies have shown that
An. koliensis populations greatly declined initially after an LLIN distribution campaign in PNG, whereas populations of
An. longirostris,
An. punctulatus s
.s.,
An. farauti s.s. did not [
13,
21]; a differential impact of LLIN that was likely due to the different host selection tendencies of the vector species. Thus, in villages like Wasab where vector diversity is high, and the species present utilize humans and other domestic vertebrate hosts for blood, the effectiveness of LLINs on malaria parasite transmission will be lessened. Such village level differences in species composition, even for those locations relatively close to each other, could explain the disparity in impact of vector control on malaria observed for a country-wide LLIN campaign in PNG [
41]. Importantly,
An.
koliensis increased proportionately in the later years of the study period in Wasab and to a lesser extent Kokofine (Fig.
2b). This phenomenon suggests a decline in the control programme’s effectiveness, particularly reduced use of LLINs, but it could also be caused by seasonal variation, which was not captured in this study.
For five of the eight
Anopheles populations analysed, the number of host-seeking mosquitoes (i.e., the bush side, unfed mosquitoes) arriving in the villages peaked between 8 pm and 10 pm in the evening and declined towards morning (Fig.
4). This relatively early arrival suggests that the adult resting sites and larval habitats for these populations were close to the villages, resulting in short commuting time between the habitats and the village. In contrast, the number of host-seeking mosquitoes in the other three populations (
An. punctulatus s.s. in Dimer;
An. koliensis and
An. longirostris in Wasab) was extended across the evening, midnight, and morning hours (Fig.
4, unfed panels) suggesting that the adult resting sites for these mosquito populations are further from the village. This observation reflects an earlier study in the north coast villages of Madang which found that blood-fed individuals of
An. farauti s.l. flew < 50 m from the study villages before resting in the nearby vegetation, whereas
An. punctulatus s.s. and
An. koliensis dispersed widely [
23]. Those researchers attributed this variation in dispersal among the species to the proximity of their preferred larval habitats and resting sites to the villages. Similarly, in East Sepik province, biting rates of
An. farauti s.l. and
An. longirostris were highest in the evening and declined towards morning, whereas those of
An. koliensis and
An. punctulatus s.s. were lowest in the evening and peaked in the morning hours [
32]. In Kokofine an early-evening biting pattern was observed for
An. farauti no. 4 [
18]. Generally, the primary host-seeking activity of most of these
Anopheles populations coincides with evening activity of villagers who would, therefore, be unprotected by LLINs.
The higher proportion of blood-fed relative to unfed mosquitoes on the village than bush side of the barrier screen and higher proportion of unfed relative to blood-fed on the bush than village side of the screen is consistent with commuting behaviour of mosquitoes. For exophilic mosquito populations like those investigated here, freshly blood-fed mosquitoes must exit the village towards their resting sites in the surrounding vegetations and are therefore likely to be intercepted by the village than bush side of the barrier screen. Similarly, unfed host-seeking mosquitoes must enter the village from their resting sites to seek vertebrate hosts and are therefore likely to be intercepted by the bush than village side of the barrier screen.
This study has two important limitations. First, the presence of the mosquito collectors near the barrier screens during mosquito collections was an unavoidable aspect of the BSS method but it also introduced potential sampling bias in favor of anthropophilic vectors. To minimize this bias, the collectors applied insect repellents on their bodies and positioned themselves further from the barrier screen and among the inhabitants of the hamlets when they were not visiting the barrier screen. Results from blood meal analysis study [
28] found that > 50% of blood-fed mosquitoes from a population of
An. farauti s.s. and a population of
An. farauti no. 4 fed on pigs or dogs. Mosquitoes in both of these populations are opportunistic feeders [
28]. As these free-roaming nonhuman hosts did not visit the barrier screens like the human collectors, this result indicates minimal effect of the collector bias; a bias result would have shown significantly more human than nonhuman blood meals. Second, the times of the year during which mosquitoes were sampled within each village was not consistent over the years (see Additional file
1: Table S1). Also, mosquitoes were not collected at the same time in the different villages. As different species of
Anopheles in PNG, particularly those within the
Anopheles punctulatus group, exhibit affinity for specific larval habitat types and the temporal distribution and abundance of these habitat types are associated with annual rainfall pattern [
40], these inconsistencies in sampling time may not capture any temporal or spatial pattern in the species composition and abundance associated with annual rainfall season. For example,
An. farauti no. 4 whose bionomics is poorly understood but believed to be associated with riverine puddles formed along the flood plains of Ramu River after a flooding event, was the only species collected in June of 2012 in Kokofine, but species richness rose to four in February 2016. Although it is possible that this change in species composition in 2016 could have resulted from a major ecological change after 2012, it is likely the result of annual seasonal variation which was not controlled for in this study.
Authors’ contributions
JBK, EDW, LJR and EKT designed the study. JBK, LJR, EKT, MK, RV, GK and NV conducted the study. EDW, LJR and SK supervised the study. JBK wrote the manuscript. EKT, LJR, SK and EDW edited the earlier drafts of the manuscript. All authors read and approved the final manuscript.