Prediction of the salmon lice density on wild sea trout from the mean predicted lice density in the sea: A cross-validation of the data in Myksvoll et al. (2018)

Summary

The coast of Norway is divided into 13 production zones (PZ 1–13) for farmed Atlantic salmon and rainbow trout. In the “Traffic Light System” (TLS) the increased mortality risk due to lice infestation on wild migrating smolt determines whether the production at the grow-out farms in a PZ must be reduced (Red light), can remain at the current level (Yellow), or can be increased (Green). Myksvoll et al. (2018) analysed observed mean lice burden per gram body weight (OMLB) of wild sea trout caught using traps and nets at 102 locations each of grid size 15 × 15 grid; i.e., 12 km × 12 km =144 km2) along the coast and the predicted mean lice particle density in the sea (PMLD) at these locations.. They concluded, based on a Spearman rank correlation of 0.72 between log of the PMLD and OMLB values, that PMLD provide reliable information about the spatial distribution of salmon lice infestation pressure all along the coast and therefore of importance for the knowledge base of TLS. Cross-validation of their data with linear regression models, bivariate logistic regression models, and a threshold model with three assumed mortality classes (Green, Yellow and Red), do not confirm their conclusions. Rather, the very low predictive power of all models shows that PMLD is such a poor predictor of OMLB that assigning a Green, Yellow or Red colour to a PZ with a sufficiently high statistical power is only possible with PMLD values from many sea areas in sum larger than the area of the majority of the 13 PZs.