In total, 14,609 blood-feeding dipterans were collected during the sampling period (Table 1), consisting of 3145 mosquitoes (64.2% females and 35.8% males) (Table 2) and 11,464 biting midges (92.8% females and 7.2% males) (Table 3). Mosquitoes were collected in 77% of all households (1–10 specimens = 73.6%; 10–100 specimens = 25.1%; 100–1000 specimens = 1.3%), and biting midges were collected in 79% of all households (1–10 specimens = 49.8%; 10–100 specimens = 39.1%; 100–1000 specimens = 10.4%; ≥ 1000 specimens = 0.7%). In total, 345 observations of mosquitoes and midges from 1035 trapping days, and 280 houses were recorded in all 42 study clusters (Table 1). Apart from sand flies [18, 19], no other haematophagous Diptera were captured in sufficient numbers to be considered in this study.
Mosquitoes (Culicidae)
Species richness
Nine genera of the Culicidae were trapped during this study (Table 2). Culex was the most abundant genus and comprised 2754 specimens (87.6% of the catches), followed by Aedeomyia (105, 3.3%), while Anopheles, Aedes and Mansonia contributed < 7% (Table 2). According to the 30% of subsampled males, Cx. quinquefasciatus was the most frequently occurring species (n = 219, 69.7%) followed by Cx. coronator (n = 34, 11.3%), Cx. bidens (n = 12, 4.0%), Cx. nigripalpus (n = 8, 2.6%), Cx. chidesteri (n = 7, 2.3%), and at least three other Culex unidentified species (n = 10, 3.3%). Culex quinquefasciatus specimens were present in similar proportions at the three trap locations (house, dog sleeping site and chicken site). The remaining specimens belonged to the group of Culex subgenus Melanoconion (n = 24; 7.6%). Aedeomyia squamipennis was the second most abundant species (n = 105, 3.3%). Within the genus Anopheles, An. triannulatus accounted for most catches (n = 14, 58.3%) and the remainder (n = 10, 42.7%) belonged to other species within the subgenus Nyssorhynchus. At least four species of Mansonia were found: Ma. humeralis, Ma. titillans, Ma. fonsecai and one unidentified species (Table 2). The genus Aedes was represented by Ae. aegypti, Ae. albopictus, Ae. serratus and another unidentified species. Three species of the genus Coquillettidia were found: Co. venezuelensis, Co. nigripalpus and another unidentified species. A few specimens of other genera were also occasionally recorded (Table 2).
Abundance and distribution in the C-arm
Mosquitoes (1372) were mostly collected at chicken roosting sites (n = 575, 41.9%), closely followed by traps located inside houses (n = 517, 37.7%) and in smaller numbers in traps near dog sleeping sites (n = 280, 20.4%) (Table 1, Fig 2a). Nine genera were recorded in chicken roosting sites, 8 in dog sleeping sites, and 7 in houses.
Annual dynamics and climatic variables
Mosquitoes were predominantly captured during the summer and early autumn (January and April 2016, rounds 16 and 17, respectively) and to a lesser extent in the early winter and spring (July and October 2015, rounds 14 and 15, respectively). The average daily temperature had a significant positive effect on the average number of mosquitoes (Z = 1.97, P = 0.049) with a 0.10 factor change per degree increase in temperature. Rainfall average did not significantly affect mosquito abundance (Z = 0.78, P = 0.437) (Fig. 3a). Up to 4× times more specimens were captured in April (the most abundant, geometric mean (GM) = 7.9; 15.0–4.8) compared to October (the poorest, GM = 2.3; 3.1–1.8). Similar annual variation was seen in all the captured genera, peaking in summer-autumn (Additional file 1: Figure S1).
Impact of insecticide interventions on mosquito abundance and distribution
Analysis of mosquito abundances (females + males) revealed a significant reduction (56%) in the PI-arm in the household traps (chicken roosting sites + dog sleeping sites + houses compared to the controls) (IRR = 0.54, 95% CI: 0.30–0.97, P = 0.04). There were significant reductions in those sites where they were most commonly caught, i.e. in houses (IRR = 0.39, 95% CI: 0.20–0.74, P ≤ 0.01) and at the chicken roosting sites, although the latter only reached borderline significance (IRR = 0.52, 95% CI: 0.25–1.07, P = 0.08) (Table 4, Fig. 2a). Analysis of female numbers alone showed that they followed a similar pattern with significant reductions overall at the household level (IRR = 0.49, CI: 0.25–0.96, P = 0.04) and also at chicken roosting sites (IRR = 0.40, 95% CI: 0.18–0.86, P = 0.02) and in houses (IRR = 0.42, 95% CI: 0.21–0.85, P = 0.02) (Additional file 2: Table S2).
The insecticidal collars did not have a significant impact on capture rates of mosquitoes at any of the household sites compared to untreated collars in the C-arm (Table 4, Fig. 2a).
Rounds 16 and 17 showed significant peaks of mosquito abundance compared to round 14 (Table 4). The number of chickens per roost was significantly associated with household level mosquito capture in all treatment arms, however, the effect was small with only a 0–2% increase in mosquito capture rate per additional chicken (IRR = 1.0, CI: 1.00–1.02, P = 0.08) (Table 4).
Biting midges (Culicoides)
Species richness
A total of 15 Culicoides species were captured. Culicoides leopoldoi was the most abundant species (n = 7057 specimens, 61.5%), followed by C. limai (n = 1877; 16.4%), and C. insignis (n = 1463; 12.8%). Small numbers of 12 other species accounted for less than 10% of the total captured (Table 3).
Abundance and distribution in the C-arm
Culicoides (n = 4803 specimens) were trapped most frequently at chicken roosting sites (n = 4066, 84.7%), followed by dog sleeping sites (n = 620, 12.9%), and to a minor extent in houses (n = 117, 2.4%) (Table 1). Thirteen species were recorded in chicken roosting sites and 11 in both dog sleeping sites and in houses.
Annual dynamics and climatic variables
Adult Culicoides were very abundant during the warmest and wettest summer sampling months (January 2016; round 16). By comparison numbers collected in autumn (April 2016; round 17), winter (July 2015; round 14) and spring (October 2016; round 15) were much lower. The average daily temperature had a positive marginal effect on the average numbers of mosquitoes (Z = 1.71, P = 0.087) with a 0.17 factor increase per degree increase in temperature. There was no significant relationship between rainfall average and Culicoides abundance (Z = 0.25, P = 0.802). The number of Culicoides trapped was much greater (14×) in January (GM = 25.2, 48.4–12.2) compared to July, which had the lowest catch (GM = 1.7, 1.9–1.3) (Fig. 3b).
Differences in abundance of the three dominant species were observed throughout the year. Culicoides leopoldoi was present in substantial numbers throughout all four sampling periods with a peak of abundance in January 2016, whereas C. limai was absent in July 2015 but present since October 2015. Culicoides insignis was particularly abundant during the rainy season (January-April 2016) but almost absent over the remaining sampling periods. The other 13 less abundant species followed a similar pattern to C. leopoldoi (Additional file 1: Figure S1).
Impact of the insecticide interventions on Culicoides abundance and distribution
Analysis of Culicoides abundance indicated that the use of lambda-cyhalothrin in the PI-arm significantly reduced (53%) the number of Culicoides (females + males) across the total of all household captures compared to the control arm (IRR = 0.47, 95% CI: 0.26–0.85, P = 0.01) (Table 4, Fig. 2b). However, when the household trap sites were examined individually, only the reduction of Culicoides in chicken roosting sites was significant (IRR = 0.48, 95% CI: 0.27–0.84, P = 0.01) (Table 4, Fig. 2b). Numbers of females alone followed a similar pattern with a significant reduction at the household level (IRR = 0.45, 95% CI: 0.25–0.81, P = 0.01) and at chicken roosting sites (IRR = 0.47, 95% CI: 0.26–0.84, P = 0.01) but not in houses or at dog sleeping sites (Additional file 2: Table S2).
The use of deltamethrin-impregnated dog collars in the DC-arm did not significantly alter Culicoides capture rates compared to untreated collars at any of the peridomestic sites (Table 4, Fig. 2b).
Rounds 15, 16 and 17 were all associated with a significant increase in Culicoides abundance compared to round 14 (Table 4). The abundance of animal hosts was a significant predictor of Culicoides capture rates, and greater numbers of both dogs and chickens were associated with larger numbers of Culicoides midges (Table 4).