Date of Award

Spring 1984

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Ocean/Earth/Atmos Sciences

Program/Concentration

Oceanography

Committee Director

Phillip R. Mundy

Committee Member

Michael J. Doviak

Committee Member

John R. McConaugha

Committee Member

Chester E. Grosch

Abstract

The reliability of an intraseason yield estimation technique which is commonly used by Pacific salmon harvest managers is evaluated for applicability to a variety of commercial finfish and crustacean fisheries. The estimation technique is known as the average timing or the average performance model. The method is not easily related to standard statistical models, but does show some similarity to both a single parameter linear regression model and the ratio estimator of sampling theory. A comparison of these models, a two parameter linear model, and a regression estimator is made to determine if the precision of forecasts of performance can be improved.

Forecasts by all methods are calculated on each successive time interval of the season. For a yield estimate by the average timing estimator, the cumulative catch of the current year is divided by the corresponding expected cumulative proportion of total yield. The time series of expected proportions is calculated from historical data. The linear model regresses annual yield on cumulative catch. Forecasts of period catches, by similar methods, have also been presented. Use of the estimation techniques has been extended to other measures of fishery performance, including catch per unit of effort (CPUE) data and abundance data.

Stratification of historical data, performed on the basis of statistical criteria, is used to select annual data series that have patterns similar to the current year. Such stratification is done in conjunction with the ratio estimator.

Six different estimators of annual performance were applied to fifty-six years of data from six different commercial fisheries. Two methods of forecasting performance for each time interval within a season were also used. The estimators were evaluated on the basis of the mean absolute percentage deviation (MAPD); where percentage deviation is the forecasting error expressed as a percentage of the forecast.

A simple linear regression model of annual performance versus cumulative performance for each time interval of the season proved to be more accurate than all other methods. In general, estimates improve as the season progresses but for all methods except the linear regression model are unreliable prior to the midpoint of the season. The overall precision of the linear regression forecasts are correlated with the variability of annual performance. Fisheries which exhibit conservative seasonal patterns of performance are well suited for this type of forecasting regime.

DOI

10.25777/hkth-2f17

Share

COinS