Date of Award

12-2008

Level of Access Assigned by Author

Campus-Only Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Marine Biology

Advisor

Yong Chen

Second Committee Member

Larry Jacobson

Third Committee Member

Huijie Xue

Abstract

Errors in modeling fishery population dynamics often come in two forms, random and structured. Random errors are common in all fisheries and have been extensively studied. Structured errors include non-random variability in biological parameters (e.g. trends in natural mortality) and non-random variability in data used to infer resource condition (e.g. trends or changes in errors for total reported catch). The former are structured "process" errors and the latter are structured "measurement" errors. A simulation study was conducted to evaluate model parameter estimation in the presence of different types of errors using an individual-based lobster simulator. The estimation model used is a size-structured, sex-specific stock assessment model for American lobster. Since the model is Bayesian, different types of likelihood functions were tested to see which yielded results with the least amount of bias and the most accuracy and precision. Both random and structured measurement errors had relatively little effect on estimation bias and accuracy; however structured measurement errors had a significant effect on estimation precision. Trends in natural mortality (M) caused a large amount of bias in parameter estimates. Neither of the likelihood functions used resulted in less bias than the other in the presence of any type of error. There was some increase in accuracy using the t distribution, but the largest effects were seen in model precision. The Student t distribution resulted in significant increases in precision in the presence of outliers and structured errors. When there was no error, or only random measurement errors, the t distribution resulted in less precision. Since lobster behavior and biology is highly affected by environmental variables, and since the model relies on survey data, a study was conducted to investigate the effect of environmental variables on the catch of lobsters in trawl surveys. Catch was found to be affected by temperature, depth, latitude, and longitude of the tow.

Share