A salmon lice prediction model

Abstract

Salmon lice (Lepeophtheirus salmonis) are parasites on salmonid fish and a density-dependent constraint to the sustainable farming of salmonids in open net pens. To control the parasites, fish farmers in Norway are required to count the number of salmon lice in different developmental stages on a subset of the fish each week. Furthermore, they must ensure that the number of adult female lice per fish does not increase beyond a specified threshold level. Here we present a model that may assist farmers in the salmon lice management. The model can predict the numbers of salmon lice in different developmental stages in each cage in a farm one to two weeks ahead. Input variables are current-week lice counts, a lice infestation pressure index, sea temperature, mean weight of the fish and presence or absence of wrasses (family Labridae) as cleaner fish. Count data for three parasitic stage groups (adult females, other motiles and sessile) are analysed jointly in one statistical model. The model predicted a large part of the variance, e.g. 50 % of the farm-level variance in adult female lice two weeks ahead. At farm-level, but not at cage-level, the numbers of other motile and sessile lice were, however, similarly well predicted by assuming "next week is the same as this week". The model also quantifies uncertainty and shows what range of outcomes is likely given the observations to that date. By using this model as decision support, fish farmers may more accurately assess the risk of exceeding lice limits.

Keywords: Bayesian model; Parasite control; Salmon lice Lepeophtheirus salmonis.