Application of New Advanced Ltering Techniques in Sh Stock Prediction

A basic sh-stock assessment system requires an integrated use of a model for the time evolution of a sh stock and information about the sh stock from catch data. Typical for common sh stock assessment systems have been the use of fairly sim-plistic data assimilation methodologies for the integration of observations with the dynamical models. On the other hand, there has been a fast development of new sophisticated assimilation techniques which can be used with highly nonlinear and complex dynamical models. In this paper some of these new methods, which are based on Markov Chain Monte Carlo formulations, will be examined with a simple sh stock model where they are shown to provide a more consistent estimate than traditional data assimilation methods. In addition, they also provide statistical error-bars in terms of error covariances for the estimates. The methods used are the ensemble Kalman lter, the ensemble Kalman smoother and the simpler ensemble smoother.