Date of Award

Fall 11-1-2020

Level of Access Assigned by Author

Open-Access Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Ecology and Environmental Sciences

Advisor

Yong Chen

Second Committee Member

David E. Hiebeler

Third Committee Member

James N. Ianelli,

Additional Committee Members

Joshua S. Stoll

Walt Golet

Abstract

The main uncertainties that affect the quality of fisheries stock assessment and pose great challenges to fisheries management can stem from a wide range of sources including observation errors associated with model input data, dynamic model process errors, model structure misspecifications, and/or volatile fishery-related socioeconomic environment. Many assessment and management failures are in part attributed to inappropriate consideration of different uncertainties. For many traditional stock assessment models, the observation error is the only source of uncertainty that modelers explicitly deal with. The observation error is usually assumed to be random. The objective functions are formulated by the corresponding distributions (e.g., log-normal distribution for biomass index data; multinomial distribution for size composition data) and minimized to estimate the model parameters (i.e., observation-error-estimators). However, the distributional assumption regarding observation error is often violated by the fact that outliers caused by atypical observation error frequently occur in fishery data. The traditional observation-error-estimators are also challenged by advocators of state-space models. Although previous simulation studies have shown the state-space model outperformed the observation-error-estimator in various cases, it is still unclear how the performance of the two estimators is affected when there are model misspecifications. In addition, despite the fact that the state-space models are advertised as providing the means to differentiate process error from observation error, the estimates of the two errors tend to be biased. How to improve the accuracy of error estimates remains a major question for state-space models. Data-limited methods (DLM) coupled with empirical harvest strategies, considered as an alternative to the stock assessment modeling approach, have been drawing extensive attention in recent decades. However, more research is needed to better understand the robustness of certain DLMs and empirical harvest strategies to key uncertainties. A biologically sustainable fishery is not warranted to be immune to socioeconomic uncertainties that may impair the wellbeing of fishermen. The exceptional slow price recovery of the American lobster in the wake of socioeconomic shocks is a living example. Analyzing the socioeconomic shocks in history can provide insights on how to help the fishery industry confront future uncertainties.

Using simulations or case studies, this project aims to evaluate the performance of existing approaches (including estimators, data processing methods, management strategies) when confronting uncertainties from different sources and develop new approaches that can better quantify the level of, or are more robust to the key uncertainties.

This study shows that using robust distributions in the likelihood function can locate outliers caused by atypical observation error in biomass index data. The advantage of the state-space production model over the observation-error-estimator diminishes with increased model specification errors. Using multiple time series instead of one time series of biomass index as model inputs can substantially improve the performance of state-space production models and especially improves the accuracy of the error estimates. A new indicator (i.e., Bhighest_S: the biomass at which surplus production is at its highest) is proposed for identifying stock status when only biomass and catch data are available. Simulation studies show that Bhighest_S is potentially more robust to observation errors or model misspecification than the modeling approach for stock status identification. Understanding the reason for the fluctuation of American lobster price suggests that providing resources gradually through the extent of price recovery rather than large and immediate injections of resources may be more efficient for fishing sectors experiencing crises.

This study provides several approaches that can better quantify or be more robust to the uncertainties commonly seen in fisheries stock assessment. While the results of the simulations and case studies are produced approximating conditions for the particular stocks (jumbo flying squid, pacific saury, American lobster), the findings regarding the uncertainty issues are relevant to many stocks that have the similar characteristics. This study evaluates error estimation in state-space production models in considerably more depth than previous studies. The recommendations made in this study can help address uncertainties in stock assessment and management.

Share