Biting midge dynamics and bluetongue transmission: a multiscale model linking catch data with climate and disease outbreaks

Abstract

Bluetongue virus (BTV) serotype 8 has been circulating in Europe since a major outbreak occurred in 2006, causing economic losses to livestock farms. The unpredictability of the biting activity of midges that transmit BTV implies difficulty in computing accurate transmission models. This study uniquely integrates field collections of midges at a range of European latitudes (in Sweden, The Netherlands, and Italy), with a multi-scale modelling approach. We inferred the environmental factors that influence the dynamics of midge catching, and then directly linked predicted midge catches to BTV transmission dynamics. Catch predictions were linked to the observed prevalence amongst sentinel cattle during the 2007 BTV outbreak in The Netherlands using a dynamic transmission model. We were able to directly infer a scaling parameter between daily midge catch predictions and the true biting rate per cow per day. Compared to biting rate per cow per day the scaling parameter was around 50% of 24 h midge catches with traps. Extending the estimated biting rate across Europe, for different seasons and years, indicated that whilst intensity of transmission is expected to vary widely from herd to herd, around 95% of naïve herds in western Europe have been at risk of sustained transmission over the last 15 years.

Figure 1

Schematic overview of the data and modelling workflow. Rectangular boxes represent data sources (solid borders denote data collected for the purpose of this study, dotted borders denote data available from literature or open-access digital archives). Circles represent models developed in this study, which have been either inferred from, or parametrised by, these data sources. Arrows denote dependency in model inference.

Figure 2

Expected midge catch size. Relative effect on midge catch size of varying mean daily temperature, mean precipitation over previous week, and catch day of year (in each case predictor variables being fixed). Left (A): Relative expected catch sizes for a range of temperatures and mean precipitation. Middle (B): Smoothed seasonal variation at the Italian sampling site if driven by observed local temperatures only (red curve) and with sine and cosine seasonality correction (black curve). Right (C): Smoothed seasonal variation at the Swedish and Dutch sampling sites if driven by observed local temperatures only (red curve) and with sine and cosine seasonality correction (black curve).

Figure 3

Diagram for midge activity model (top) with midge catch counts and catch model predictions (bottom). Black circles show catch counts and orange crosses are the median prediction of the GLMM over all possible random effects, with orange lines connecting the median predictions of catch weeks. Background shading show GLMM catch predictions between 5 and 95% percentiles for random effect coefficients, with deeper shading indicating a collection week. Images created by Viola Visser.

Figure 4

Predicted daily $R_p$ dynamics for farms in the area of the trap sites in each country (shown in insets). Left column (A): ${\mathit{R}}_{\mathit{p}}$ on each day of year (March-November) for P = 5% (blue curve), median P = 50% (black curve) and P = 95% farms (red curve) averaged over each year’s prediction for that day. Individual years (2000–2015) are shown as fainter curves of same colour. Right column (B): The mean ${\mathit{R}}_{50}$ over all days March-November for each year 2000–2015 (grey bars) with the maximum ${\mathit{R}}_{50}$ value, over all days March-November for each year 2000–2015 (blue circles).

Figure 5

Estimated proportion of naive herds at risk of BTV outbreaks ($P$). Estimations over time periods of five years and early (March–May), mid (June–August) and late (September–November) seasons mapped across Europe for 2000–2015 (spatial cell scale = 0.25° lat./long.). Colours indicate increasing proportion of risk from yellow to red (zero risk is coloured grey). It should be noted that the risk map was calibrated using serological data for BTV-8 serotype and midge capture data that did not include C. imicola (see Results and Discussion).

Figure 6

Schematic representation of the cattle herd level BTV transmission model. The population of cattle and infected biting midges are divided amongst discrete disease compartments. BTV latent midges (E_M) enter the model at a rate proportional to the daily prediction of the catch model. The extrinsic incubation period for latent midges is modelled as a multi-stage process before midges become infectious (I_M). Susceptible cattle (S_C) become infectious cattle (I_C) after a bite from infectious midges (I_M), to become resistant cattle (R_C) in time (also modelled as a multi-stage process; red box). Transitions are shown as solid lines, coloured according to their dependence on environmental variables: constant per-capita (black), daily mean temperature dependent (red), all predictor variables of capture model and the catch-to-bite scale parameter $\mathit{\xi }$ (blue). Dotted lines indicate where the number of infected individuals in one species increases the incidence rate in the other species. Outcomes of the model are linked to observed cattle milk serology time series by the first two infectious stages for cattle with virus being undetectable by ELISA (blue box), whereas subsequent infectious stages and the recovered stage are detectable by ELISA (green box). The likelihood function for $\xi$ was inferred by marginalisation over the latent stochastic variables affecting model outcomes (e.g. herd-specific random effects, daily fluctuations in midge activity). Used images were available under open licence Creative Commons Deed CC0.