Statistical modelling of Sea Lice counting data from salmon farms in the Faroe Islands (Oral presentation, 2016)

The Faroese salmon industry is troubled by high sea lice levels leading to expensive treatments against sea lice. Fiskaaling is regularly counting the number of sea lice in the different development stages (chalimus, mobiles and adult) for the salmon farms in the Faroe Islands. A statistical model of the counting data is developed in order to understand and simulate 1) the adult sea lice levels, 2) the ineffective treatments and to 3) help reduce the sea lice levels. In the model, the sea lice infection is represented by the chalimus lice developing into adult lice. For Faroese waters, the development times of about 1-2 months were inferred from EWOS growth-curves for sea lice as a function of temperature. The chalimus counts are used to simulate past and current levels of adult lice, and to predict the adult sea lice level 1-2 months into the future. The modelling was done with R/RStudio and plots made with the ggplot2-package. Error bars for the counting data were Poisson 95% confidence-intervals made with poisson.exact() from the package exactci. The performance of the model is demonstrated by simulations of 6-12 months long time series of adult sea lice counts for different farming sites and years in the Faroe Islands. In most cases, the observed and simulated data are not significantly different, and the ineffective treatments can also be understood and simulated. Time series analysis of the counting data using ccf() from the xts-package is also applied. Significant lagged correlations are found between the time series of chalimus and adult lice. Although site variability exists, a preliminary analysis from 7 sites indicates that the median time delay (48 days) is consistent with the EWOS development times (median 46 days). This method may possibly prove useful for independent checks and in-situ measurements of the sea-lice dynamics.