Record Details
Evaluating the data-poor fishery stock assessment method, DB-SRA
ScholarsArchive at Oregon State University
| Field | Value |
|---|---|
| Title | Evaluating the data-poor fishery stock assessment method, DB-SRA |
| Names |
Owashi, Brandon R.
(creator) Sampson, David B. (advisor) |
| Date Issued | 2014-09-02 (iso8601) |
| Note | Graduation date: 2015 |
| Abstract | Depletion-Based Stock Reduction Analysis (DB-SRA; Dick and MacCall, 2011) is a catch-only fisheries stock assessment model that has been developed to estimate an overfishing limit (OFL) in data-poor situations. DB-SRA projects the biomass trajectories of a stock by means of a catch time series and five parameters: the instantaneous, per annum, rate of natural mortality (M), age at 50% maturity, F[subscript MSY]/M, B[subscriptMSY]/B₀, and the predicted depletion of the stock from its unfished condition. F[subscriptMSY]/M is the rate of fishing mortality associated with the maximum sustainable yield (MSY) divided by the natural mortality rate, and B[subscriptMSY]/B₀ is the biomass level associated with the MSY divided by the unfished level of biomass. DB-SRA performs a Monte Carlo simulation where a large number of random parameter draws are made based on the input parameter’s prior distribution. Based on the catch time series, a biomass trajectory is produced to estimate a feasible set of input parameters and an OFL. The run and corresponding set of input parameters are not retained if the biomass trajectory goes below zero. In instances where the input parameter prior distributions are unknown, Dick and MacCall (2011) proposed a set of default values for two life history types (rockfish and flatfish). Although DB-SRA has been evaluated to some extent and is currently being used for management of data-poor species on the U.S. west coast, further evaluation is warranted. Like other fisheries assessment models, DB-SRA makes several assumptions that may have large influence on outputs and have largely gone untested. First, in essence DB-SRA assumes that only mature fish are caught in the fishery; this is rarely true on the U.S. West Coast and elsewhere, particularly for species with substantial recreational catch. Second, most stock assessment methods, including DB-SRA, are applied to large regions (e.g., U.S. west coast), assuming the population dynamics and fishing behavior remain consistent across the entire area. Market demands and habitat, among other factors, can lead to heterogeneity in population dynamics and fishing behavior. For instance, immature fish are often caught in recreational fisheries, but commercial fisheries tend to target larger fish, causing fishing impact to change across regions. I developed a two-region operating model that simulated data to generate input parameter expected values and a catch time series for each region, then conducted a factorial experiment to investigate the effect of four factors on DB-SRA (version 4) results: (1) different positions of the selectivity curve (the relative vulnerability to fishing of each age class) relative to the maturity curve; (2) spatial scale (separate by region versus combined); (3) exploitation history; and (4) life history type (rockfish and flatfish). The position of the selectivity curve influences the accuracy of the OFL estimates from DB-SRA, whereas the exploitation history has little effect. The OFL estimates are overestimated when the selectivity curve is to the right of the maturity curve and underestimated when the selectivity curve is to the left of the maturity curve. DB-SRA produces higher OFL estimates when two regions are used instead of one large region. Dividing the catch data into multiple regions resulted in higher OFL estimates than one combined region when the same input parameters and catch time series were used. An updated production function, for mimicking population dynamics, was implemented in DB-SRA (version 4), creating separate time lags for mortality and production (recruitment and growth). Instead of setting the time lags for mortality and production equal to the age at 50% maturity (version 3), the time lag for mortality has been changed to one year (version 4). Although the version 4 DB-SRA model has been used for fishery management, it has not been formally evaluated against version 3 to understand the impacts of this change on model results. To investigate the two versions, I looked at different positions of the selectivity curve relative to the maturity curve, different exploitation histories, and varying spatial scale for two life history types. The OFL estimates from version 3 of DB-SRA were larger than the OFL estimates from version 4, which is also evident in the biomass trajectories. The biomass trajectories from version 3 are always greater than the respective biomass trajectories from version 4. Although the OFL estimates from version 4 are not always less biased than those from version 3, the estimates from version 4 are always more precautionary and significantly reduce the chances for overestimating the OFL. The identification of factors that influence DB-SRA OFL estimates could demonstrate how DB-SRA can be adjusted to produce less biased OFL estimates in more situations. The change made in the production function between versions 3 and 4 of DB-SRA makes OFL estimates more precautionary; but does not always reduce the bias in the median OFL estimate. The results from this study could provide information to fisheries managers so that DB-SRA could be potentially improved and is applied in appropriate situations. |
| Genre | Thesis/Dissertation |
| Access Condition | http://creativecommons.org/licenses/by/3.0/us/ |
| Topic | stock assessment |
| Identifier | http://hdl.handle.net/1957/52387 |
