We investigate how cod feeding data may be included into a recently developed age-structured model for a coastal cod population in a small fjord in northern Norway. The model integrates data from mark-recapture experiments, trawl-surveys and commercial catches. The fjord population has high biomass density (2-3 tonnes km-2) and low fishing mortality rates (F <0.30 year-1). The model output parameters (cannibalism mortality rate, other natural mortality rate, fishing mortality rate, recruitment numbers, gear selection functions for commercial and research trawls and trawl catchability coefficient) are estimated using a weighted least-square minimisation routine. We evaluate different strategies for, and effects of, inclusion of feeding data in the model, and evaluate the effects of uncertainty in the different data sets. Representation of the feeding data as an Ursin lognormal distributed size (predator weight/prey weight) food preference function yielded realistic cannibalism mortalities. The sensitivity of the results to data input are discussed. |