Data assimilation into biomass dynamics models : a Monte Carlo simulation experiment

Abstract

In this paper, we use a variational data assimilation method to fit biomass dynamics models to simulated data. The method is the variational adjoint technique in which a cost function measuring the distance between the data and the model solution is minimized. This approach is a deterministic procedure in which the model is repeatedly solved and the solution compared to the observations or measurements in order to find the parameters of the model that give predictions which are as close as possible to the data. We will briefly review some of the methods commonly used in fisheries management and compare them with the method in this paper. Twin experiments are used to evaluate the performance of the algorithm. The parameter estimates have reflected the true values well and the deviations from the true parameters are not substantial compared with the errors in the simulated data.