The Research On Performances Of Surplus Production Models Under Different Fishing Scenarios

The paper compares the most popularly used surplus production models using simulated data and swordfish, Xiphias gladius, fishery data in the North Atlantic Ocean. None of the four models could work well in the scenario 2, under explored stock, however they got accurate results in scenario 1, regulated stock. The Schnute model work well in full explored stocks such as scenario 1, 3, 4 and 5, especially in low biomass stocks such as scenario 3, but not suitable to the under explored stocks such as scenario 2. The W-H model generally performed well in all kind of scenarios especially in scenario 1, 4, and 5. In the swordfish, xiphias gladius, fishery, estimation of W-H model and D-Fox model produced MSY about 14000 ton, very near the result of Prager in 1996.The author tentatively explored the suitable fishing condition for the models to stock assessment. To make reasonable estimation biologically, the fishery population must experience over explored stage and stock recovery stage.In generally, the more complex regression models were, the higher regressionr2 they could be. In examining parameter estimates from any fitted assessment model, an analyst must always question the validity of the results biologically. A new appraisal criterion was proposed to question the validity of the results, the reliableindex (re). This index, based on the coverage index (C) of Prager(1996), not only presents the estimatabilty of observed fishery data, but also expresses the risk of the models results. Analysis of 6 fishery scenarios showed that the re index could correctly indicate the validity of the models results; however the coverage index could not work well when regression models made blunder.In analysis the 6 fishery data sets year by year, it was found that, there werestrong correlation among the estimated parameters: the estimated r , q correlated positively, and the estimated r , K and q, K correlated negatively.