The impact of temperature changes on vector-borne disease transmission: Culicoides midges and bluetongue virus

Abstract

It is a long recognized fact that climatic variations, especially temperature, affect the life history of biting insects. This is particularly important when considering vector-borne diseases, especially in temperate regions where climatic fluctuations are large. In general, it has been found that most biological processes occur at a faster rate at higher temperatures, although not all processes change in the same manner. This differential response to temperature, often considered as a trade-off between onward transmission and vector life expectancy, leads to the total transmission potential of an infected vector being maximized at intermediate temperatures. Here we go beyond the concept of a static optimal temperature, and mathematically model how realistic temperature variation impacts transmission dynamics. We use bluetongue virus (BTV), under UK temperatures and transmitted by Culicoides midges, as a well-studied example where temperature fluctuations play a major role. We first consider an optimal temperature profile that maximizes transmission, and show that this is characterized by a warm day to maximize biting followed by cooler weather to maximize vector life expectancy. This understanding can then be related to recorded representative temperature patterns for England, the UK region which has experienced BTV cases, allowing us to infer historical transmissibility of BTV, as well as using forecasts of climate change to predict future transmissibility. Our results show that when BTV first invaded northern Europe in 2006 the cumulative transmission intensity was higher than any point in the last 50 years, although with climate change such high risks are the expected norm by 2050. Such predictions would indicate that regular BTV epizootics should be expected in the UK in the future.

Keywords: European epizootic outbreaks; climate change; vector-borne disease in northern Palaearctic.

Figure 1.

Temperature-dependent rates and capacity per vector. Temperature-dependent rates for midges: biting rate per day (a), mortality rate per day (b), rate of BTV incubation within the midge vector per day (c). (d) Vectorial capacity per vector at static constant temperatures for the three EIP models. The saw-tooth pattern for deterministic (fixed duration) EIP is due to modelling bites at discrete daily activity periods rather than assuming that midges can bite continuously. For each EIP model, transmission was maximized by a static temperature of 19.2–19.6°C; the intensity of transmission was greater for models where variance in EIP is larger.

Figure 2.

First 20 days of maximizing temperature sequences for Ĉ. The daily sequence of temperatures that imply maximal vectorial capacity per vector (Ĉ*) for each of the three EIP models: deterministic (a), fitted to BTV EIP in midges (b) and Markovian (c). The bars indicate how much the temperature would need to vary on that day to reduce Ĉ by 1% compared with Ĉ*. The dashed lines show the optimal static temperature for Ĉ*_static.

Figure 3.

Daily temperatures, expected midge biting and BTV risk for 2006. (a) The daily mean temperature for 2006 (CET dataset). (b) Expected number of midge bites per cattle host for each day in 2006 from the susceptible midge population. (c) The vectorial capacity per vector on each day (Ĉ(t)) of 2006 for deterministic EIP (black curve), BTV-fitted EIP (red curve) and Markov EIP (blue curve). Static temperature capacity per vector with deterministic EIP also given (grey curve). (d) The seasonal reproductive ratio on each day (R(t)) of 2006 for deterministic EIP (black curve), BTV-fitted EIP (red curve) and Markov EIP (blue curve). Static temperature reproductive ratio (R_static(t)) per vector with deterministic EIP also given (grey curve).

Figure 4.

Trends in yearly cumulative risk and average R(t) profiles. (a,c,e) The cumulative reproductive ratio R_cum for each historical year from 1950 to 2015 (bars) and UKCP09-derived realizations for 2016–2049 (dots) using the deterministic (a,b), BTV fitted (c,d) and Markov (e,f) EIP models. Three periods are highlighted: baseline 1961–1990 years (green), recent years 2000–2015 (purple) and midterm future years 2035–2049 (orange). Other years are given in grey. The 10 year running averages are shown (dashed curves). (b,d,f) Daily averages for reproductive ratio (〈R(t)〉) for baseline (green), recent (purple) and future (orange) for each EIP model. Future years were also averaged over UKCP09 realizations for each year. Individual year R(t) profiles are shown as thinner curves. The length of the epidemic season is when R(t) > 1 (black dashed line).