Variance partition reveals contrasting random effect contributions on the density and species composition of malaria-transmitting mosquitoes

Spatial-temporal variations exist in the density and species composition of malaria-carrying mosquitoes, which will in turn influence the transmission of the disease. While there has been extensive research on the seasonality and other main drivers of the vector populations, the heterogeneity induced by random effects is just as important but did not quite attract the same attention. To investigate the relative contributions of the between-house, between-village, and between-year variations, as well as other house-level covariates such as inhabitant number and bed net usage, intensive Pyrethroid Spray Catches (PSC) sampling was conducted across a 60-month period between 2012-2019 from four villages in the Sudano-Sahelian region of Burkina Faso. For mosquito density, measured by counts of female Anopheles gambiae s.l., our modelling showed that the between-house variation was the largest component, followed by the between-year then between-village variation, after accounting for seasonality and other covariates. Density increased with the number of inhabitants within a household but was uncorrelated with bed net usage. A subset of female mosquitoes was genotyped for species identification, and the composition of An. coluzzii and An. gambiae, the two dominant vectors in the region, varied hugely across villages without a clear seasonal trend. The between-village variance contributed up to 76% of the total random variation, followed by the between-year variance. The between-house variation was estimated to be statistically insignificant. Neither household size nor bed net usage had any impact on species composition. In short, the relative importance of the random components in mosquito density was in the reverse order from species composition. The estimates and relative strengths help parameterise potential field trials for novel vector control programmes and monitoring.