Maximum midge abundance and the risk of bluetongue Adam Butler1, Beth Purse2, Gustaf Rydevik3, Kate Searle2 1Biomathematics and Statistics Scotland, Edinburgh, Scotland 2Center for Ecology and Hydrology, Easter Bush, Scotland/ Wallingford, England 3Roslin Insitute, Edinburgh University, Scotland Contact: gustaf.rydevik@roslin.ed.ac.uk Bluetongue is a viral livestock disease that is spread by midges (in particular midge species from the Culicoides taxa), which causes severe illness and often death in sheep, and somewhat milder symptoms in cattle. While it is an endemic disease in countries surrounding the equator, the UK has only experienced one outbreak of bluetongue, in 2007, that had a severe economic impact. There has recently been a number of outbreaks of BTV in France, causing concern for the potential of a new UK introduction. This work attempts to predict the potential number of new cases generated if an infected midge was introduced at different locations in Europe, including the UK Midge abundance Bluetongue spread We used Culicoides surveillance data to develop a modeling framework for predicting the maximum annual abundance, taking into account the effect of environmental covariates (e.g. precipitation, land cover type, host density, amount of vegetation,…)  on the maximum abundance of Culicoides at each site. For each site and year, we then used the maximum size of weekly female catch per species per year as a response variable. The model can be described as Max_abundance (at site i in year t) ~ Poisson(mu[i,t]) Log(mu[i,t]) = a + bi*xi + site(random effect) + individual_level(random effect) Once fitted using INLA, predictions were generated for locations that were comparable and for which we had training data. Predicted maximum midge abundance of Cullocoides In order to estimate the potential number of new cases (the R0), we made use of a mathematical formula published by Simon Gubbins in 2008. Gubbins set up a set of differential equation describing the spread of disease as a function of the expected number of infected midges generated by an infected host (which depends on midge density), the probability of a infected midge surviving long enough to infect a new host (which depends on temperature) and the expected number of hosts (sheep and cattle) bitten and infected by a single infected midge. The expression above is the eigenvalue of these equations. This equation was used with the median posterior estimate of maximum midge abundance, and combined it with datasets on summer surface temperature and cattle densities to calculate a surface of R0 values. Distribution of R0 for bluetongue in Europe Areas in red and yellow are those with an R0>1 These are very preliminary results, and there are currently a few odd things about them. For the midge modelling aspect, we have excluded some locations where the predicted abundance were unrealistically high (more than several million midges), but we don't yet understand what was causing the model to misbehave. A second issue is that the calculated R0 values are somewhat to low to be realistic - we know from the 2007 outbreak that southern England is suitable for spread, and yet the estimated values are predominantly lower than 1. Methodologically, a challenging issue is related to temperature; we are modelling the max abundance, which only occur at a single point in time, but we don't model when it will occur, so it is not obvious what period to use for calculating the temperature used in the R0 model. Finally, we are currently simplifying the statistical model by simply assuming that the yearly maximum is poisson distributed. Modelling the max abundance explicitely by using the extreme value likelihoods implemented in INLA might lead to a better model.  Issues and further work