In fisheries, advice for the management of fish populations is based upon management quantities that are estimated by stock assessment models. Fisheries stock assessment is a process in which data collected from a fish population are used to generate a model which enables the effects of fishing on a stock to be quantified. This study determined the effects of various data sources, assumptions, error scenarios and sample sizes on the accuracy with which the age-structured production model and the Schaefer model (assessment models) were able to estimate key management quantities for a fish resource similar to the Cape hakes (Merluccius capensis and M. paradoxus). An age-structured production model was used as the operating model to simulate hypothetical fish resource population dynamics for which management quantities could be determined by the assessment models. Different stocks were simulated with various harvest rate histories. These harvest rates produced Downhill trip data, where harvest rates increase over time until the resource is close to collapse, and Good contrast data, where the harvest rate increases over time until the resource is at less than half of it’s exploitable biomass, and then it decreases allowing the resource to rebuild. The accuracy of the assessment models were determined when data were drawn from the operating model with various combinations of error. The age-structured production model was more accurate at estimating maximum sustainable yield, maximum sustainable yield level and the maximum sustainable yield ratio. The Schaefer model gave more accurate estimates of Depletion and Total Allowable Catch. While the assessment models were able to estimate management quantities using Downhill trip data, the estimates improved significantly when the models were tuned with Good contrast data. When autocorrelation in the spawner-recruit curve was not accounted for by the deterministic assessment model, inaccuracy in parameter estimates were high. The assessment model management quantities were not greatly affected by multinomial ageing error in the catch-at-age matrices at a sample size of 5000 otoliths. Assessment model estimates were closer to their true values when log-normal error were assumed in the catch-at-age matrix, even when the true underlying error were multinomial. However, the multinomial had smaller coefficients of variation at all sample sizes, between 1000 and 10000, of otoliths aged. It was recommended that the assessment model is chosen based on the management quantity of interest. When the underlying error is multinomial, the weighted log-normal likelihood function should be used in the catch-at-age matrix to obtain accurate parameter estimates. However, the multinomial likelihood should be used to minimise the coefficient of variation. Investigation into correcting for autocorrelation in the stock-recruitment relationship should be carried out, as it had a large effect on the accuracy of management quantities.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:rhodes/vital:5571 |
| Date | January 2008 |
| Creators | Northrop, Amanda Rosalind |
| Publisher | Rhodes University, Faculty of Science, Statistics |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis, Masters, MSc |
| Format | 93 p., pdf |
| Rights | Northrop, Amanda Rosalind |
Page generated in 0.0027 seconds