The Shockwè trap: a human-baited exposure-free device for surveillance and behaviour studies of anthropophilic vectors

Background: The human biting rate (MBR) and entomological inoculation rate (EIR) are common parameters routinely used to measure the risk of malaria transmission. Both parameters can be estimated using human landing catches (HLC). Although it is considered the gold-standard, HLC puts collectors at higher risk of infection with mosquito-transmitted pathogens. Methods: A novel exposure-free host-seeking mosquito electrocution trap, the Shockwè trap (SHK), was developed and its efficiency for monitoring mosquito community composition and abundance was compared with human landing catches (HLC) as the gold-standard. Field experiments were performed in Massavasse village, southern Mozambique. Simultaneous indoor and outdoor collections of nocturnal host-seeking mosquitoes were carried out using the SHK and HLC methods. The relative sampling efficiency of SHK was estimated as the ratio of the numbers of mosquitoes caught in SHK compared HLC. Proportionality and density-dependence between SHK and HLC catches were estimated by mean of Bayesian regression approaches. Results: A total of 69,758 and 27,359 host-seeking mosquitoes comprising nineteen species and four genera, were collected by HLC and SHK respectively. In general, SHK and HLC sampled similar numbers of mosquito species, with the exceptions of the least common species Aedes sudanensis, Ae. subargenteus, and Coquillettidia versicolor that were caught only by HLC. The relative sampling efficiency and proportionality between SHK and matched HLC catches varied greatly between species and collection site. However, all mosquitoes collected by SHK were unfed, confirming the Shockwè trap design’s performance and reliability as a successful mosquito exposure free sampling approach. Conclusions: Results demonstrate that SHK is a safe and reliable human-exposure free device for monitoring the occurrence of a wide range of mosquito, including major malaria and arboviruses vector species. However, improvements are needed to increase its sampling efficiency for less abundant mosquito species.


Introduction
The magnitude of a human host's exposure to biting by the mosquito vector is a key parameter in epidemiological studies of malaria.This biting exposure is still determined using the human landing catch (HLC), a method first introduced in the 1930s 1 , but that continues to be widely used and accepted as the gold standard method for measuring biting exposure.A simple procedure requiring minimal equipment, HLC also provides a reliable method for measuring diurnal patterns of vector activity, and for identifying body regions most frequently contacted or preferentially bitten by a particular species 2 .However, HLC has serious operational drawbacks: it is physically demanding and its accuracy greatly depends on the individual collector's stamina, honesty, experience, their inherent attractiveness to mosquitoes [3][4][5] and their ability to catch them, especially at high biting rates 6 .Collectors may face a higher risk of disease transmission by the mosquitoes being studied or by other species, particularly the mosquito vectors of arboviruses against which there are no effective vaccines, like dengue, Zika and chikungunya.This may not always be true, because, at least in some circumstances, there is good evidence that, with malaria, there is negligible risk to HLC collectors 7,8 .Hence, a longstanding goal for medical entomologists and vector epidemiologists has been the development of a sampling device that can accurately reproduce the entomological indicator obtained with HLC without posing risk to collectors.
Typically, traps for sampling host-seeking anthropophilic mosquito vectors either incorporate a live human bait or are designed to incorporate a human host mimic assembled from visual or olfactory cues.The yield of traps with synthetic baits can be dramatically improved using carbon dioxide, but the expense and logistics of using and controlling this highly volatile mosquito stimulant/attractant, has severely limited its effective use in the majority of routine applications to date.Moreover, the physiological effect of several synthetic odour compound blends remains poorly defined, as the level of attractiveness can dramatically vary depending on the vector species 9,10 .Moreover, some commonly used attractant blends, as well as CO 2 can function as repellents at certain concentrations 11 .
A wide range of traps exist, and many innovative designs continue to be developed today [12][13][14][15][16][17][18][19][20] .A number of studies have evaluated and compared some of the more popular traps [21][22][23][24] , and although there is no single trap design that is consistently reported as being an improvement on HLC, the simple, compact, and relatively inexpensive CDC light trap is very widely used for sampling and surveillance of nocturnally active mosquitoes worldwide.A major advantage of light traps is that sampling can be conducted across large areas in short time periods, with minimal effort.However, light traps also have limitations.For example, many species of vector mosquitoes routinely sampled by light traps do not naturally orient to light, raising questions over what the mosquitoes are responding to the trap's light or whether they are simply being captured passively as they drift close to the trap's fan.Moreover, light trap collections are greatly compromised by environmental conditions, particularly moonlight 6,25,26 , to an extent where outdoor collections are not directly comparable with indoor collections performed simultaneously 6,27 .
Here we report on the development and field testing of a novel trap for collecting mosquitoes attracted to a sleeping human bait, a device with potential for use indoors and outdoors.The trap exploits the knowledge that Anopheles gambiae s.l. and other mosquitoes preferentially descend onto a supine human host 28,29 , the sleeping position presented to hungry host seeking female anophelines and other nocturnally active mosquitoes by the vast majority of humans.The trap can be used by a single person (or with another source of attractant), who remains free from exposure to mosquito bites for the entire sampling period, while the mosquito responds to natural olfactory and physical stimulants produced by the host.

Ethical considerations
The study received ethical approval from the To minimise the risk of infection, all volunteers (for both HLC and SHK) during the field studies were provided with anti-malaria prophylaxis, Fansidar ® (Sulfadoxine-Pyrimethamine), in accordance with the recommendations of the Mozambique National Malaria Control Program at the time the study was undertaken.A recent study showed that this practice reduced risk of malaria infection in HLC volunteers by 96.6% compared to a matched control group of individuals not involved in the study but living in the same locality 7 .

Consent
Written informed consent was provided by all volunteers participating in the study, by signing a form that was co-signed by the senior investigator.

Description of study site
The study was conducted during the wet season February to April 2016 in Massavasse village (-24.624839°S; 33.111787° E).The village is located in Chókwè district, southwest Gaza Province, Mozambique) 30 (Figure 1).The mean temperature in this period typically ranges from 25-34°C, and this is also the season experiencing most of the annual rainfall of 600 mm.Anopheles arabiensis, An. funestus, An. gambiae s.s, An. tenebrosus, An. pharoensis and An.ziemanni are the most common malaria vectors in the village.Several other mosquito species within the genera Culex and Aedes known to be vectors of arboviral diseases have been also found in the village 30,31 .

The Shockwè trap
The Shockwè trap (SHK) (Figure 2) comprised a metal frame, 200 cm in length, 100 cm wide at the base, 70 cm wide at the top and 65 cm in height from the trap base to the roof (equivalent  to the maximum flight height reported for several Anopheles species 32,33 ).The base of the trap was 10 cm from the floor, compatible with the reported flight height of several mosquito species [34][35][36] .A modified insect electrocution grid device (Bower Products, London;), delivering 3,800 volts at 9 milliamperes (input of 230 volts) at the electric grid, was deployed on top of the frame.The device was fully compliant with EU safety specifications for use (BS EN 60335F2F59 and the European EMC directive).The electrocution grid measured approximately, 65 × 68 cm (16 cm deep) and the area of the active electric net was 493 × 500 mm.A grounded outer aluminium mesh of a size that allowed mosquito entry but prevented any human body parts from contacting the grid protected the live wires.The electric grid was positioned on the frame above the head and torso region of a supine adult lying beneath, as this is where the majority of both Anopheles (and Culex spp.) approach the host at the top surface of the net 28,37 (Figure 2).The human bait slept below the electrocution grid inside an untreated bed net trap which had been modified to function as a mosquito collecting sack (Figure 2).The electricity supply incorporated a circuit breaker (a residual current device or RCD), and in the event of a short circuit, the circuit would cut out instantly, significantly reducing the risk of serious injury.In the field, the trap was powered by a 7.5 kW 4-stroke Ryobi™ generator, providing an uninterrupted alternate electric current (AC) supply to the traps for up to 11.50 hours without refuelling.

Experimental design
The experiments were conducted in five of Massavasse's six neighbourhoods.A series of paired indoor and outdoor collections were conducted in six selected sentinel sites located at the five selected neighbourhoods (Figure 2).In each site, four experienced adult collectors were randomly assigned to perform paired mosquito collections indoor and outdoor by mean of HLC and using SHK trap.All collectors were trained adult males, above 18 years old.Indoor collections were made using a purpose-built portable experimental hut, described previously 30 .The relative positions of the paired collectors performing HLC and SHK were located 12 to 20 metres apart to prevent any possible interference between them.Collection was performed from 19:00 to 05:00 hrs, and collectors performed hourly rotation between collection methods to minimize the effect of different attractiveness of different collectors or their catching ability on the quality of collected data.

Sample processing and analysis
Collected mosquito samples were immobilised within paper cups inside a refrigerator (approx.4°C) for 20 to 30 minutes.Anopheline mosquito species were identified morphologically to species or species complex level (e.g., An. gambiae s.l.) using the taxonomic keys of Gillies & De Meillon 38 and Gillies & Coetzee 39 .Culicine mosquitoes were identified using taxonomic keys of Edwards 40 , Jupp 41 , Harbach 42 , and Service 43 .Members of the major vector species complexes, such as An.gambiae s.l or vector species groups (e.g., An. funestus s.l) were identified to species level by molecular analysis (PCR).The protocols of Scott et al. 44 and Koekemoer et al. 45 were used to detect species-specific single nucleotide polymorphisms (SNPs) in the intergenic spacer region (IGS) and internal transcribed spacer region 2 (ITS2) of ribosomal DNA (rDNA) genes for identifying members of An. gambiae complex and An.funestus group, respectively 44,45 .

DNA extraction
Whole mosquitoes were transferred into 1.5 ml Eppendorf tubes and submerged with 200 µl of Tris-EDTA buffer solution at pH 7.8 (Sigma-Aldrich ® , USA).The samples were then macerated until homogenize and incubated at 94°C for 12 minutes, after which the homogenized were centrifuged at 13,000 rpm for 4 minutes for deposition of fragments of exoskeleton tissues.Then, 100 µL of the supernatant (solution containing DNA) were transferred to a new pre-coded Eppendorf tube, and store at -20°C for further analysis.

DNA amplification
Amplification of extracted DNA sample was performed by mean of conventional PCR.The primers and reagents for master mix used are indicated in Table 1.The primers were obtained from Bei Resources (https://www.beiresources.org).1.2 µL of extracted DNA were added to solutions containing the master mix.For identification of members of An. gambiae complex, PCR reactions were run with the following condition: one initial denaturing cycle at 94°C for 2 minutes.Thirty denaturing, annealing and extension cycles at 94°C, 50°C and 72°C, respectively, all for thirty seconds, a final extension cycle at 72°C for 5 minutes.Reaction products were hold at 4°C inside thermocycle (Biometra ® , USA) before visualization using electrophorese gel 44 , mixed with 10ul of SybrSafe (ThermoFisher TM , USA).For An. funestus group, the following programmer were used: one initial denaturing cycle at 94°C for 2 minutes.Thirty denaturing, annealing and extension cycles at 94°C, 45°C and 72°C, respectively, all for thirty-six seconds, and a final extension cycle at 72°C for 5 minutes, and a holding cycle at 4°C for indefinite time until visualization 45 .For An. gambiae s.l members, the reactions creates DNA fragments of 464Pb (An.merus, ME), 390Pb (An.gambiae, GA), 315Pb (An.arabiensis, AR), 153Pb (An.quadrianulatus, QD), 153Pb (An.quadrianulatus, QDA).Likewise, for An.funestus group the primers create DNA fragments of 600Pb (An.vaneedine, VAN), 464Pb (An.funestus s.s, FUN), 4111Pb (An.rivolurum, RIV), 4111Pb (An.rivolurum Like), 252 Pb (An.parensis, PAR) and 146Pb, 166Pb (An.leesoni, LEES) 44 .

Statistical analysis
Estimation of sampling efficiency.Commonly used methods for analyzing agreement between measures 46 are not suitable for mosquito count data as counts are usually very skewed and highly over dispersed.In addition, the usual approach of log transforming counts to reduce the degree of skewness is not applicable because of the frequency of counts of zero 47 .However, mosquito counts in candidate sampling devices can be ranked, and agreement with matched HLC quantified using rank concordance correlation coefficients (CCC).CCC statistics quantify the variability in the data but do not give any indication of whether the novel method is collecting information that is comparable with the standard approach.Estimates of the trapping efficiency can be computed by taking the ratios of total numbers of mosquitoes caught by the different methods, but these are very sensitive to outliers in the counts, either of the novel or the standard methods 47 .
In view of these constraints, to determine SHK trap sampling efficiency relative to HLC, we used a less outlier-sensitive regression-based approach, that had been originally proposed by Hii et al., 48 , and was adopted subsequently by other authors 17,27,49 .Comparisons were made between indoor HLC, outdoor HLC, indoor SHK trap and outdoor SHK trap.Strata were defined by combinations of sampling locations and days.The number of strata included in the analysis varied by mosquito species taxon.
The following linear statistical model was used to estimate the sampling efficiencies of SHK relative to HLC, ( ) ( ) where: E(y i ) is the expected number of mosquitoes of a given taxon caught using method m in stratum i; E(x i ) is the expected number of mosquitoes of the same taxon caught using an indoor HLC method in the same stratum i; α m is the relative sampling efficacy for method m, compared to indoor HLC for which the value is set to unity.

Species Reagents Qty
Anopheles gambiae s.l.GoTaq DNA polymerase (5 U/ul) 0.1 μL assumed in the observed numbers of mosquitoes caught by each methods so that: x i ~Poisson(E(x i,m )) and: y i,m ~Poisson(E(y i,m )) and the model therefore assumes the distribution of the numbers of mosquitoes caught by any method to be a log-normal mixture of Poisson distributions.
To examine whether the trapping efficiency varied with the average mosquito density, the following extended model was also fitted: where γ m is an exponent corresponding to method m.
If the different methods are sampling the same fraction of the mosquito population, then the fitted line for model (1) should be close to that for model (2).Equivalently, the 95% credible intervals for y m should overlap with unity.A value of y m that is different from unity indicates a lack of proportionality between the mosquito sampling methods.In addition, α' m will differ from α m if γ m is different from unity 48 .The model was fitted using a Bayesian Markov chain Monte Carlo algorithm in rjags v. 4 -14 50 .

Abundance and composition of mosquito collections
A total of 35 nights of paired HLC and SHK collections yielded a total catch of 97,117 mosquitoes comprising twentythree species in five genera (Table 2).Sampling with HLC and SHK yielded 69,758 and 27,359 specimens respectively, with Culex and Anopheles being the most common genera sampled by both methods.
Anopheles species comprised Anopheles gambiae s.l, An. pharoensis, An. tenebrosus, An. ziemanni and An.funestus, and accounted for 17.8% (12,401 n = 13,853) of the total (indoor and outdoor) Anopheles species catches obtained by HLC and 5.3% (n = 1,452) sampled by SHK.Culex tritaeniorynchus was the predominant species amongst eight Culex species sampled by either method (Table 2).These species represented  3).A total of 126 and 60 specimens from HLC and SHK respectively, initially identified morphologically as members of An. gambiae complex did not amplify.

Shockwè trap efficiency
Rank concordance correlation analyses between HLC and SHK catches are summarized in Table 3.For most species, estimates of CCC suggested moderate association between SHK and matched HLC catches (Table 4).Regression model analyses indicated that trapping efficiency of SHK varied in function of mosquito species.Indoor and outdoor, the result indicated that SHK was efficient at sampling Ae. fryeri and An.pharoensis (Figure 3A, Table 5).In addition, the trap was also efficient in sampling Cx. bitaeniorhynchus, Cx. tritaeniorhynchus and Ma.uniformis outdoor (Figure 3B; Table 5).
The ratio of total caught outdoor to indoor suggests that these mosquito species were primarily exophagic (Table 5).
Regression model parameters showing the degree of proportionality between SHK and HLC catches, and density-dependency of SHK performance are depicted in Table 6.For most of species, regression models indicated that the number of mosquitoes collected by SHK were density dependent, the 95% credible interval for the exponent (γ m ) did not overlapped with unity (Table 6).The exception was observed with An. gambiae s.l (indoor), Cx pipiens s.l (indoor) and Ma.uniformis (indoor) (Table 6, Figure 4).For other species that SHK and HLC collections were at some point proportional, such as An.funestus s.l (indoor and outdoor), An. tenebrosus (indoor and outdoor), An. ziemanni (indoor and outdoor), Ae. fryeri (indoor), the large 95% credible interval indicates huge uncertainty (Table 6).

Discussion
While the SHK proved a sensitive device for detection of all but the rarest of the full range of species in the study site, it collected far fewer mosquitoes than the HLC caught for most species.This is not an uncommon feature of many traps, and as with the CDC light trap and other traps, does not necessarily mean they are unsuitable for use as epidemiological tools 48,52,53 .Mosquito species detected by SHK and HLC included dominant malaria and arboviruses vector species.Unfortunately, it has yet to be determined how or why the HLC method outperforms numerous traps.The skill of the collectors, often considered a potential limitation of HLC at high densities, did not appear to impair the numbers of individuals captured at the high densities recorded in the present study, while variations in the inherent attractiveness to mosquitoes of different hosts would have influenced performance of the SHK to a similar extent as the HLC.Despite their importance in epidemiological studies, and their decades-long use for surveillance of vector mosquito populations of public health interest, knowledge of how mosquitoes behave at close range to a trap or of how the integration of visual and olfactory stimuli enable mosquitoes to find a stationary host is limited [54][55][56][57] .Uniquely, the SHK is designed to capture mosquitoes as they descend onto a supine human host in response to the rising body heat and associated   attractants, a host-location route previously identified from infra-red tracking studies as the most common route taken by An. gambiae s.l and An.arabiensis when approaching a human host 28,58 .Assuming the SHK was operating as intended, any other species arriving at the host in this way should also have struck the electrocution grid and been killed and captured.This would include Cx quinquefasciatus 37 but the orientation paths of the other important vector species, have not been characterised.However, horizontal air movements across the host would disperse the plume of attractants and eliminate the focus of attractants at the net roof 59 and the SHK would become very ineffective.As such conditions would be common outdoors but rare indoors the SHK would be expected to be more suitable for use indoors than outdoors.
SHK sampling efficiency was, in general, density-dependent suggesting that SHK trap could fail to detect uncommon or rare mosquito species.Several factors, such as trapping location, visual contrast, trap configuration and, as well as climate and surrounding environment features determine trap response of insects during appetitive and attraction flights 60,61 .
The extent to which climate factors, notably air temperature, relative humidity, and wind speed influence the sampling efficiency of traps still remains poorly known.These factors may influence trap performance by affecting vectors appetitive searching for host clues [62][63][64][65] , whilst wind speed may directly affect efficiency by reducing or fully discouraging flight activity.
Mosquitoes typically fly at a cruising velocity less than 1 m/s so that, wind speed above normal flight speed may completely cease flight activity of several vector species 66,67 .During this study air temperature at night was within the optimum range for mosquito flight activity (20°C -30°C), as reported in another study also conducted in southern Mozambique 68 and elsewhere 63,69 .Differently, the average wind speed at sunset and, usually at dawn, was around 4.15 m/s, above of what is considered optimum speed for mosquito to fly 67 .One of the main advantages of SHK is the fact it is fully humanbaited, and in this first evaluation, all the catches comprised unfed malaria and arboviruses vector species, notably An. arabiensis, An. gambiae s.s, An. funestus s.s., An. merus, An. parensis, Cx. quinquefasciatus, Cx. tritaeniorhynchus and Ma.uniformis, attesting to its exposure-free.Here, the SHK provided a reliable sample of anthropophilic mosquito species that arrive unfed at the host with a genuine intention of blood-feeding on the volunteer within the bed net.Furthermore, the trap offers an alternative to explore in real time key behavioural components concerning host-finding behaviour by disease vectors using recently developed image-capturing technology such as those reported by Parker et al., 28 .However, it is worth noting that, although both SHK and HLC employ the same source of odour stimuli (natural human host), it is very unlike that they could, thereby, collect absolutely the same quantity of mosquitoes since both are subjected to imprecisions and analytical errors that can generate significant levels of variability 46,70,71 .However, as Hii et al., 48 have remarked, the lower numbers of mosquitoes caught by an alternative method do not per se invalidate the use of the candidate method for estimation of relevant entomological indicators.The crucial criterion is whether the new method is collecting mosquitoes that are in proportion to the gold standard.This condition was fulfilled with An. gambiae s.l, Cx. pipiens s.l and Ma.uniformis, but not with remaining vector species.
One of the key limitations of the SHK trap is dependence on a source of alternating electric current (AC) for operating.However, this limitation could be reduced by installing an AC/DC power convertor or by modifying the electrocution grids to enable them to work with direct current (DC).It was noted that the voltage input would sometimes inflict significant burn damage on mosquitoes, particularly on small-sized mosquitoes such as An.funestus rendering them difficult or even impossible to identify morphologically.However, identifying the ideal settings to eliminate this risk may is not a simple or predictable process and finding the optimal level of voltage is likely to be a major challenge.Setting the voltage was compromised by the need to ensure that the live wires comprising the electrocution grid were spaced at a distance to eliminate short-circuiting, but close enough to capture even the smallest anophelines and retaining enough voltage to knock them down, without damaging them beyond recognition.

Conclusions
In conclusion, the Shockwè trap is a human-baited exposure-free mosquito trapping device with high potential as a sampling device for the indoor and outdoor entomological surveillance of malaria and arbovirus vectors.It also could serve as an experimental device designed to capture mosquito species as they respond to a supine human host.Despite their essential role in elucidation of the mechanism of host location by tsetse, electrocution grids have yet to considered as tools suitable for use in similar studies on mosquitoes 72 .Structural improvements are possible and could improve trapping at low densities, a challenge for all similar mosquito traps.

Data availability Underlying data
Open Science Framework: Underlying data for 'The Shockwè trap: a human-baited exposure-free device for surveillance and behaviour studies of anthropophilic vectors' https://doi.org/10.17605/OSF.IO/6K4MT 51 This project contains the following underlying data: • Dataset_1: Raw dataset of mosquito species collected using human-landing catch and Shockwè trap in Massavasse village from February to April 2016. addressed.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound?Yes Are sufficient details of methods and analysis provided to allow replication by others?Yes

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility?Yes

Are the conclusions drawn adequately supported by the results? Partly
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Vector biology, vector surveillance I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
mosquito-transmitted pathogens.While it is true that there is a risk of exposure during the HLC, there is evidence suggesting that if the collectors are monitored correctly and provided with prophylaxis, the risk of exposure may not be higher than those not involved in the process, and in some conditions, it may even be lower.

Specific comments
Although the HLC method collected 2.5 times more mosquitoes than the SHK method, the author's conclusion favoured SHK by giving more importance to the number of species collected by each method.Why was the number of species collected preferred over the number of mosquitoes collected for comparison?○ It's worth noting that the SHK trap effectively catches most species of mosquitoes without exposing humans to infection.However, the author did not discuss the inherent limitations of the SHK trap in capturing the most important malaria vector, the An.gambiae complex, and its impact on vector surveillance and monitoring for decision-making.Why this point was not discussed?

○
The authors did not focus much on discussing the practical implementation of the trap for collecting mosquitoes in rural areas where there is no electricity, despite the fact that mosquito monitoring and surveillance are mainly carried out in these areas.Why wasn't the practicality of SHK traps in malaria-endemic rural areas discussed compared to HLC? ○ Why did the authors favour SHK traps despite the limitations in collecting the number of mosquitoes and their practicality in areas that require reliable monitoring tools?○ Additionally, why did the authors tend to SHK traps without concrete evidence indicating that the risk of human exposure is higher during the HLC method?The mosquito control program requires efficient and practical tools to provide suitable data for decision-making with little or no risk of exposure.Reviewer Expertise: Public health entomology I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Figure 1 .
Figure 1.Location of Gaza Province, southern Mozambique showing the location of Lionde administrative post where the study site at Massavasse is situated.

Figure 2 .
Figure 2. The Shockwè trap (SHK) prototype showing (A) SHK as deployed outdoors in use in the field; (B) a side view of a volunteer within the trap and a top view of the collection sack or bag, which, when in use, is positioned beneath the electrocution grid; (C) the dimensions and structure forming the trap frame, and (D) a typical sample of mosquitoes from one SHK trap on one night collection using in Massavasse village.

Figure 3 .
Figure 3.The efficiency of the Shockwè trap (SHK) trap deployed indoors (A) and outdoors (B) in Massavasse village.In each case the vertical dashed line corresponds to a trapping efficiency of 1.0 (indicating nondifference).The error bars are 95% credible intervals.

Consent
Written informed consent for publication of the participants' details was obtained from the participants.

Figure 4 .
Figure 4. Proportionality between Shockwè trap (SHK) and human landing catch (HLC) indoor catches of Anopheles gambiae s.l, Culex pipiens and Ma.uniformis.In each case, the dashed diagonal line corresponds to a 1:1 ratio of the two methods (trapping efficiency 1.0).The green straight lines are the fitted lines from the linear models.The blue curves correspond to the fitted lines from the power model and the blue shading to the 95% credible interval around it.

○
Is the work clearly and accurately presented and does it cite the current literature?YesIs the study design appropriate and is the work technically sound?YesAre sufficient details of methods and analysis provided to allow replication by others?YesIf applicable, is the statistical analysis and its interpretation appropriate?YesAre all the source data underlying the results available to ensure full reproducibility?YesAre the conclusions drawn adequately supported by the results?PartlyCompeting Interests: No competing interests were disclosed.
Research Ethics Committee at Liverpool School of Tropical Medicine (Research Protocol 14.055, Host Location by Exophagic African Malaria Vectors) on 28 th January 2015 and from the Comité Nacional de Bioética para Saúde (CNBS) of the Ministry of Health of Mozambique (MISAU), Ref: 208/CNBS/15, on 22 nd July 2015.

Table 1 . Reagents and primers used for preparation of master mixes for molecular identification of member of An. gambiae complex 44 and An. funestus group 45 .
The underlying mosquito density E(x i ) is assumed to have a log-normal distribution, i.e.

Table 2 . Relative abundance and composition of mosquito collections by each sampling methods: Human landing catch (HLC) and Shockwè Trap (SHK); deployed indoors or outdoors in Massavasse village.
51 (n = 22,078) and 16.1% (n = 4,401) of specimens sampled by HLC and SHK, respectively.The genus Mansonia was represented by two species, Ma. africana and Ma.uniformis which accounted for 50.4% (n = 35,183) and 78.1% (n = 21,361) of all specimens sampled by HLC and SHK, respectively.Two Aedes mosquito species, namely Ae. (Stegomyia) subargenteus and Ae (Muscidus) sudanensis, and one Coquillettidia species were only collected by HLC (Table2).The underlying human landing catch and Shockwè trap raw data can be found here (https://doi.org/10.17605/OSF.IO/U5A4V)51.Molecular identification of species complex/group members Of 1,052 mosquitoes subjected to molecular analysis, 980 were identified as members of the An.gambiae complex and 72 as in the An.funestus group (Table3).Of the 980 Anopheles gambiae s.l, 850 were collected by HLC and 176 by the SHK 701 were identified as Anopheles arabiensis, 20 as An.gambiae s.s, 68 as An.merus and 26 as An.quadriannulatus.
Of 31Anopheles funestus group processed, 22 were An.funestus s.s and 7 were An.parensis.All the members of An. gambiae s.l and An.funestus identified by PCR were detected in the subsamples from both SHK and HLC, with the exception of An. merus and An.parensis, which were collected only with HLC (Table

Table 4 . Total mosquitoes caught in Massavasse village, total strata, and concordance correlation coefficients.
Indoor human landing catch (HLC) collections were arbitrarily considered as reference for concordance analysis.SHK: Shockwè Trap.

Table 5 . Estimates of trapping efficiency (α) from linear models (95% credible intervals
). Indoor human landing catch (HLC) collections were arbitrarily considered as reference in regression models.Only species that occurred on at least ten sampling occasions (strata) were included in the models.SHK: Shockwè Trap; DIC: Deviance information criterion.

Table 6 . Parameter estimates from power models (95% credible intervals) applied to investigate density-dependence in sampling performance of Shockwè trap (SHK) compared with human landing catch (HLC).
Indoor HLC collections were arbitrarily considered as reference in regression models.Only species that occurred on at least ten sampling occasions (strata) were included in the models.
*Results unreliable due to sparse data