link to AFSC home page
Mobile users can use the Site Map to access the principal pages

link to AFSC home page link to NMFS home page link to NOAA home page

NOAA Technical Memorandum NMFS-AFSC-302

Publications Overview
Pubs Database
New Publications
Poster Presentations
Processed Reports
Quarterly Report:
Current Issue
Archives
Index
Feature Articles
Feature Archives
RACE Cruise Archives
Reports to Industry
Stock Assessments
Tech Memos
Yearly Lists

Deployment performance review of the 2014 North Pacific Groundfish and Halibut Observer Program

Abstract

This report contains the analyses and findings of the Alaska Fisheries Science Center’s Fisheries Monitoring and Analysis Division’s Observer Science Committee on the efficiency and effectiveness of observer deployment in 2014. This review is performed as part of the annual review process implemented by the North Pacific Fishery Management Council for the North Pacific Groundfish and Halibut Observer Program. This report is intended to inform the Council, the National Marine Fisheries Service (NMFS), and the public how well various aspects of the observer program are working, and it contains recommendations for improvement.

Several performance metrics were evaluated. These included metrics that assess whether targeted levels of observer coverage were achieved, metrics that assess whether data from the observed fleet was representative of the unobserved fleet, and metrics that evaluated the probability of obtaining data throughout the fleet given the current sample rate (deployment rate). Observer deployments are representative if they are not spatially or temporally biased and if the characteristics of observed trips are similar to those from unobserved trips. Non-representative spatial or temporal deployments are evidence of a deployment effect and differences between the observed and unobserved fleet fishing characteristics are evidence of an observer effect.

This evaluation of observer deployment was conducted at the stratum level as defined in the 2014 Annual Deployment Plan (ADP; released in December of 2013), since each stratum is defined by its own sampling unit (i.e., vessels or trips) and/or rate of sampling. Representativeness of trip selection was examined at a finer scale to help clarify initial findings.

In 2014, observers were deployed and collected data for 4,367 days. This amount is within 9% of the observer days budgeted at the start of the year in the 2014 ADP. In total, 417 vessels and 5,883 trips (43%) were observed in the fleet from 11 deployment strata. Both full coverage strata were completely observed, the coverage rate in the trip selection stratum was within its expected range, the coverage rate within the six vessel selection strata were within expected ranges (4 strata) or greater than expected amounts of coverage (2 strata), and there was no coverage in the no selection strata (2 strata).

In trip selection, vessels log trips into the Observer Declare and Deploy System (ODDS) where logged trips are randomly selected for observer coverage based on pre-determined selection probabilities. Exact binomial tests confirmed that the ODDS was selecting trips at the correct rate (16%). However trips selected to be observed were cancelled at nearly four (3.7) times the rate of unselected trips. In order to prevent cancellation of observed trips from decreasing the sample size, the ODDS was constructed to automatically assign observer coverage to the user’s next logged trip; however, the order and timing of observed trips can be changed by the user. The lag in the number of observed trips relative to the number of total trips in this stratum was evidence of temporal bias in trip selection. There was no evidence of spatial bias in this stratum. We recommend that the ODDS protocols of 1) allowing selected trips to be cancelled and 2) allowing multiple trips to be logged prior to sailing be re-evaluated.
           
There was evidence of an observer effect in trip selection. Initial tests found differences in the vessel length and landed catch of observed trips compared to unobserved trips. When differences were examined by gear, it was found that hook-and-line vessels landed 14.4% less catch that had 9.1% more landed species when observed than when unobserved. Trawl vessels were found to have fished in 4.4% fewer areas and fished on trips that were 8.4% shorter in duration when observed compared to when unobserved. We recommend Fisheries Management Plan (FMP) area and gear type be included as potential strata definitions in the 2016 ADP.
           
In vessel selection, NMFS randomly selects some vessels for 100% coverage during six separate 2-month selection periods (strata). Sampling frames for each strata are constructed based on past fishing activity. This method has three drawbacks. First, vessels that fished in the past are not necessarily those that will fish in the future. Second, vessels that did not fish in the past but do fish in the selection period have no chance of being selected. The error in terms of the percentage of the selection frame represented by these two factors averaged 40% among selection periods. Third, vessel selections need to be inflated since some selected vessels will request and be granted a release from observer coverage by NMFS. The number of selected vessels was on average 8 times higher than the targeted number of vessels in 2014.

Releases from coverage in the vessel selection strata occurred in each selection period in 2014. Between 37% and 67% of vessels that were selected and fished in a given time period were granted a release from observer coverage by NMFS. There was no evidence of spatial bias caused by releasing vessels in those areas where more than 10 trips occurred; however, four of the areas with fewer than 10 trips were not observed due to releases and there were more trips released from coverage than were observed. However, vessel selection coverage rates remained within or above expected values (23% of vessel selection vessels were observed during 2014). There was strong evidence of an observer effect in vessel selection; the number of permutations tests with p-values < 0.05 was 9 times the expected amount indicating fishing characteristics differed between observed and unobserved trips.
           
Based in part on the recommendations contained within the 2013 version of this document, the vessel selection method of observer deployment was discontinued in 2015. Vessels previously in vessel selection now participate in trip selection. To increase sampling efficiency in future years, we recommend that a list of vessels that cannot carry a human observer be generated and updated each year. This list could then be used to define new deployment strata for non-human monitoring in future ADPs.
           
Observers monitor the deliveries of trawl pollock at shoreside processing plants and collect genetic tissue samples from salmon bycatch. For vessels fishing pollock in the full-coverage stratum, this task is performed by designated plant observers, whereas for vessels fishing pollock in partial-coverage strata, only those trips that are observed at sea are also monitored at the plant. Although all pollock deliveries were monitored in the full coverage pollock fishery, this was not the case for pollock fishery trips in the partial coverage stratum where the proportion of deliveries monitored for salmon bycatch was lower than the target sample rate. Observer coverage rates were especially low in ports with high tendering activity. When tendering activity was removed from the analysis, the proportion of pollock fishery partial-coverage deliveries that were monitored fell within the expected range, indicating salmon sampling goals were not achieved in the partial coverage pollock fleet as a result of tendering activities.

Tendering activity is problematic for the observer program since observers rely on the captain to define the fishery, there are insufficient resources in most partial-coverage plants to reliably detect salmon bycatch if they are rare, and observers cannot be stationed on tendering vessels. Vessels that delivered to a tender were 11.5% shorter in length, fished 29.1% longer in duration, and had catch that was 1.3% less “pure” than vessels that did not deliver to a tender. Because of these differences in trip characteristics when a vessel was delivering to a tender versus delivering shoreside, an examination of potential observer effects was conducted within trips delivering to tender vessels. Trips delivering to tenders occurred on vessels that were 8.8% shorter in length and delivered catch that was 6% less “pure” when observed relative to unobserved trips, providing evidence of an observer effect within the fleet of vessels delivering to tenders. The benefits to monitoring salmon bycatch in the presence of deployment and observer effects caused by tendering activity should be addressed in future ADPs.
           
We examined the probability of having no observer coverage as a function of how many trips were in a NMFS area for each partial coverage stratum since each stratum has its own coverage rate. The probability of having no data from a NMFS area decreased as total effort and sampling rate increased. While this result is not surprising, it helps to highlight that the smaller the population of interest (e.g., fishing activities in a NMFS area), the higher the overall coverage rate needs to be in order to meet observer coverage goals and to obtain data that are representative of fishing activities.
           
Catcher processors (CPs) in the partial coverage stratum pose a sampling problem to the observer program. Since there are few trips made by these vessels, the probability that they are observed under the current selection rates is low. In addition, by nature of being CPs, fishing trip characteristics tend to be different from the other trips in partial coverage. With this in mind, we recommend that catcher processor vessels in partial coverage be treated as a new stratum with a potentially different selection rate.

We repeat our 2013 recommendation that NMFS should develop a trip identifier to better link databases containing logged trips that are used in the logistics of observer deployment (ODDS) to the actual landings records of fishing trips (eLandings).


View Online  (.pdf, 6.7 MB).
 


            | Home | Site Map | Contact Us | FOIA | Privacy | Disclaimer | USA.gov | Accessibility | Print |           doc logo