In this paper a catch projection model for annual forecasting of Brazilian sardine (Sardinella brasiliensis) production in the state of Santa Catarina is proposed. A time series of catch (metric tons), fishing effort (landings) and mean monthly yields (tons/landings) between 2000 and 2005 was used to stipulate observed scenarios likely to be repeated in the forthcoming years. A linear model was fitted to the variables yield and month for each year and a systematic re-sampling technique was applied to choose the final scenarios (slopes) from the models with smallest residuals. Monthly and weekly projections of sardine production were conducted for 2006 according with the scenarios stipulated in the previous years. For all the projections the model has shown satisfactory forecasting and a significant error reduction towards the end of the fishing season. The levels of precision obtained in 2006 suggest that the forecast model presented can be a useful tool for the management process of this fishery and the mediation of current conflicts between fisherman and the fish processing industry.