The Study Between Continuous Schaefer-Form Production Model And BP Network Model In Fishery

In fish stock assessment,traditional models have three main types,there are production models,depletion models,and age-structured models.With the development of computer technology,geographic information system,remote sensing technology and artificial neural networks technology gradually applied to the assessment and prediction of fisheries resources.Surplus production models are major academic models in assessment and management of modern fishery resources.Surplus production models can be simply used in fishery assessment because of their simplicity and relatively undemanding data needs.Surplus-production models generally do not incorporate age structure which is useful for fish population dynamics.These models are of particular value when the catch can not be aged,or can not be aged precisely.Moreover, Surplus-production models are classical methods to predict the MSY(maximum sustainable yield) for fisheries management,which is one of the major fisheries management goals.BP neural network model is emerging in recent years,because of its high degree of parallelism,adaptive self-leaming function,it has people's attention.In this study,the author using the Culter alburnus resources in DanJiangkou Reservoir and the annual fish catch in East China Sea as an example,to compare Schaefer model and BP model,the contents and main results as follows:(1) Using the continuous Schaefer model and C.alburnus resources in DanJiangkou Reservoir to estimate B₁(initial biomass),r(innate rate of increase),K (environmental carrying capacity),q(quotiety of catch),MSY(maximum sustainable fishing yield) and f_MSY.The results are:B₁=186302 kg,K=2200429 kg,q=0.2123, MSY=113216 kg,f_MSY=458172.(2) Using the continuous Schaefer model to estimate the total catch in East China.Sea,results are:B₁=12236567 t,K=29546790,q=0.398,MSY=4109876 tk f_MSY=13.63.(3) Using the BP neural network model to build an yield prediction model for DanJiangkou Reservoir,using the data from 1981 to 1985 as indicate of comparison, the results are:predictive value from 1981 to 1985 followed by 108150 kg,119339 kg, 126845 kg,114010 kg,109652 kg.(4) Using the BP neural network model to build an yield prediction model for East China Sea,using the data from 1991 to 1995 as indicate of comparison,the results are:predictive value from 1991 to 1995 followed by 4094768 t,4709258 t, 5349900 t,6349296 t,7314227t.(5) Compare the results' accuracy of assessment using Schaefer model and BP model in different fisheries.Deviation of parameters are: in DanJiangkou Reservoir,B₁:9.7,r:2.7,K:3.9,q:2.3,MSY:1.6,f_MSY:1.0; in East China Sea:B₁:17,r:8.8,K:12,q:7.5,MSY:1.7,f_MSY:1.3. The error rate between predictive value and true value are:in DanJiangkou Reservoir, from 1981 to 1985 followed by 0.0672,0.0621,0.0503,0.0341,0.0485;in East China Sea:from 1991 to 1995 followed by 0.0040,0.0521,0.0745,0.0736,0.0767.