Reference Document Home | Publications Home
CONTENTS
Introduction
Data Sources
Methods
Results & Discussion
References
Acknowledgments
Appendix A
Appendix B

Northeast Fisheries Science Center Reference Document 08-02

A Brief Description of the Discard Estimation for the National Bycatch Report

S.E. Wigley, M.C. Palmer, J. Blaylock and P.J. Rago
National Marine Fisheries Serv, Woods Hole Lab, 166 Water St, Woods Hole MA 02543-1026

Web version posted January 29, 2008

Citation: Wigley SE, Palmer MC, Blaylock J, Rago PJ. 2008. A brief description of the discard estimation for the National Bycatch Report. US Dept Commer, Northeast Fish Sci Cent Ref Doc. 08-02; 35 p.

Information Quality Act Compliance: In accordance with section 515 of Public Law 106-554, the Northeast Fisheries Science Center completed both technical and policy reviews for this report. These predissemination reviews are on file at the NEFSC Editorial Office.

Acrobat Download complete PDF/print version

INTRODUCTION

NOAA Fisheries is currently preparing a National Bycatch Report summarizing estimates of discards, by species, which occurred in 2005 in all federally managed fisheries in the United States.  This document briefly describes the methods used to estimate the discards of finfish and shellfish in 2005 in fisheries in the Northeast Region, which will be included in the National Bycatch Report.  The regional analysis involved 33 species and 77 fleets (Table 1 and Table 2).  Stock components were not considered in the analyses, and only fleets for which discard estimates could actually be derived will be included in the National Bycatch Report.

The discard estimation process used a stratification approach broad enough to encompass all species, and employed a combined ratio method using a discard-to-kept weight ratio.  The discard estimates derived will not necessarily directly correspond with those contained in individual stock assessments due to differences in stratification and estimation methods.  However, the various estimates should be of the same order of magnitude.


DATA SOURCES

Northeast Fisheries Observer Program Data

Northeast Fisheries Observer Program (NEFOP) data were used to calculate discard ratios.  Only observed hauls from observer trips in 2005 for which a ‘complete’ sampling protocol were analyzed.  Training trips, aborted trips, and hauls with no catch reported were eliminated from the data set.  Species hail weights with discard reason ‘039’ (previously discarded) were also excluded.  Conversion factors were applied to convert any dressed weight data to live weight equivalents.   Observer trips were assigned to fleet sectors using the NEFOP program codes:  

NEFOP program code

Fleet sector

130

US/CAN Resource Sharing area

140

Haddock longline Hook sector

150

B-day

201-204

Scallop access areas

000

Open area

Vessel Trip Report Data (VTR)   

As dealer records in the Northeast do not contain information on mesh size and area fished of the vessel trips involved in the purchases, these data could not be used to expand the observer discard ratios, by species and fleet, to calculate total discards.  However, this information is recorded on Northeast Vessel Trip Reports (VTRs); thus, the VTR data were used to expand the NEFOP discard ratios to total discards.  In the analysis, all of the commercial VTR trips in the 2005 database were used (excluding NY state [non-federal] vessels).  As with the observer data, conversion factors were applied to convert various units of catch to pounds, live weight.

Surfclam Logbook and Dealer Data

The surfclam fishery has its own separate logbook system (different from the VTR system).  As such, the data in the 2005 surfclam logbooks were used to augment the 2005 VTR data for the surfclam dredge fishery. 

Days-At-Sea Data

VTR fishing trips were assigned to a fishery sector using the 2005 Days-At-Sea (DAS) database.  It was assumed that all vessels that fished under a Special Area Access Program reported their participation (as required) either by the Days-At-Sea Call-In System or via the Vessel Monitoring System (VMS).  The DAS database integrates both systems.  Five access area classifications were used: ‘closed area,’ ‘US/CAN resource sharing area,’ ‘B-day program,’ ‘hook sector,’ and ‘open area.’  If a fishing trip was not assigned to one of the first four access area categories, it was assigned to the ‘open area’ category.


METHODS

In all of the analyses, the sampling unit was an individual fishing trip.  Trips were partitioned into fleet sectors using six classification variables: calendar quarter, area fished, gear type, mesh size, access area, and trip category.  Calendar quarter was based on the landed date of the fishing trip, and was used to capture seasonal variations in both fishing activity and discard rates.  Area fished was based on statistical reporting area; trips where area fished was not recorded or was otherwise unknown were excluded.  Two regional areas were defined:  New England (NE) comprising statistical reporting areas <‘600’ (which includes Southern New England, Georges Bank, and the Gulf of Maine), and Mid-Atlantic (MA) comprising statistical areas >=‘600’.  Gear type was based on Northeast gear codes (negear).  Some gear codes were combined into a single category (Table 2), and trips for which the gear was unknown were excluded.  Mesh size groups were separately created for otter trawl and gillnet gear.  For otter trawls, two mesh groups were formed:  small mesh (less than 5.5 inches) and large mesh (5.5 inches and greater).  For gillnet, three mesh groups were formed: small mesh (less than 5.5 inches); large mesh (between 5.5 and 7.99 inches); and extra large mesh (8 inches and greater).  Five access area categories were used: ‘closed area,’ ‘US/CAN,’ ‘B-day,’ ‘Hook,’ and ‘open area.’ Sea scallop fishing trips were divided into General (Gen) and Limited (Lim) category trips.

DAS data (fishery codes, DAS codes, and access area codes) were used to assign all VTR trips into one of five access area categories.  Vessel permit number and date landed were used to link VTR trips with the DAS trips.  A detailed description of the methods developed (and obstacles encountered) to link the VTR and DAS databases is provided in Appendix B.

When one or no observer trips occurred in a calendar quarter, an imputation approach was employed to ‘fill in’ the missing (or incomplete) information using data from an adjoining stratum.  In this imputation procedure, only the temporal stratification (i.e., calendar quarter) was relaxed to half year, recognizing that seasonal variations occur for some species.  When all quarterly cells were missing for a fleet, or sparse observer coverage existed across all quarter for the fleet, the fleet was subsequently eliminated from the analysis.

Discard Estimation

Total annual discards were estimated using a combined d/k ratio estimator (Cochran 1963) where d = discard pounds of a given species, and k = the kept pounds of all species.  Total discards (in weight) of a species by a fleet were derived by multiplying the estimated discard rate for that particular species in that fleet by the corresponding fleet landings in the 2005 VTR database.

The combined ratio method is based on a ratio estimate pooled over all strata and all trips within a fleet.

The total discard (in pounds) of species j was defined as:

(1)   Equation 1   

where

(2) Equation 2       

where

Dj hat is the total discarded pounds of species j;
Kh is the VTR total kept pounds in stratum h;
rc,j is the combined ratio of species j;
djih is the total discards (in pounds) of species j in trip i in stratum h;
kih is the kept pounds of all species on trip i in stratum h;
Nh is the number of VTR trips in stratum h; and
nh is the number of observed trips in stratum h.

In Equation 2, the summation over strata h = 1 to Q occurs over calendar quarters.  Equation 3 (below) requires a more explicit definition of the stratum designation as the summation over quarters relies on the annual combined ratio defined in Equation 2.

The variance of Dj hat for species j was defined as:

(3) Equation 3     

where

Dj hat is the total discards (in pounds) of species j;
Kqh is the VTR total kept pounds in quarter q and stratum h;
rc,j is the combined ratio of species j;
djiqh is the total discards (in pounds) of species j in trip i in quarter q and stratum h;
kiqh is the kept pounds of all species on trip i in quarter q and stratum h;
Nqh is the number of VTR trips in quarter q and stratum h; and
nqh is the number of observed trips in quarter q and stratum h.

The coefficient of variation (CV) of Dj hat was defined as:

(4)   Equation 4

All discards were assumed to result in 100% mortality.  If survival ratios are used in a stock assessment, then a survival ratio are applied to the discard estimates presented here.  Survival ratios are available for spiny dogfish and summer flounder (Appendix A Table A1).

Method Validation

Validation of the approach used to estimate total discards was performed by using this same approach to estimate the landings of each of the species in 2005, and comparing these estimates to the landings included in the VTR and Dealer databases.

To estimate landings using the NEFOP data, the same estimation method was used; however, the species-specific poundage discarded (dj) was replaced with species-specific kept pounds (kj).

 (5)   Equation 5   

where

(6) Equation 6       

where

Lj hatis total kept pounds of species j;
Kh is the VTR total kept pounds in stratum h;
rc,j is the combined ratio of species j;
kjih is the total kept pounds of species j in trip i in stratum h;
kih is the kept pounds of all species on trip i in stratum h;
Nh is the number of VTR trips in stratum h; and
nh is the number of observed trips in stratum h.

In Equation 6, the summation over strata h = 1 to Q occurs over calendar quarters. Equation 7 (below) requires a more explicit definition of the stratum designation as the summation over quarters relies on an annual combined ratio defined in Equation 6.

The variance of Lj hat for species j was defined as:

(7) Equation 7   

where

Lj hat is the total kept pounds of species j;
Kqh is the VTR total kept pounds in quarter q and stratum h;
rc,j is the combined ratio of species j;
kjiqh is the kept pounds of species j in trip i in quarter q and stratum h;
kiqh is the kept pounds of all species on trip i in quarter q and stratum h
Nqh is the number of VTR trips in quarter q and stratum h; and
nqh is the number of observed trips in quarter q and stratum h.

The coefficient of variation of Lj hatwas defined as:

(8)   Equation 8

For each species, 95% confidence intervals were calculated for the point estimate of total landings.

RESULTS AND DISCUSSION

Using the 2005 observer data, discards were estimated for 33 species in 25 of the 77 fleets examined (Table 1, Table 2, and Table 5).  A total of 3,565 trips[1] were observed in 2005, with the majority of these occurring in the otter trawl, gillnet, and sea scallop dredge fleets.  Although observer coverage in 2005 was relatively high compared to previous years, some fleets had little or no observer coverage (Table 2).  For some fleets with limited temporal coverage by observers, imputation was used to derive the discard estimates.  However, using half-year estimates may not be appropriate for all species and, in some cases, quarterly discard ratios were based on very small sample sizes.  This contributed to the lower precision (higher CVs) associated with several of these estimates (Table 6).

The 2005 VTR landings (all species combined, live weight), by fleet and quarter (Table 3), were used to expand the discard ratios (Table 4).  Total discards in 2005 (in metric tons), by species and fleet—with and without survival ratios applied— are presented in Tables 5a and 5b, respectively.  Because discards were not estimated for all fisheries (due to data limitations), the values in Table 5 underestimate the actual total discards in 2005.

Qualitative comparisons of the 2005 discard estimates (using both the annual totals and the totals for specific gear) with other recent discard estimates available for the same species indicated a similarity in order of magnitude.  That is, the 2005 estimates approximate those derived from: (a) the Standardized Bycatch Reporting Methodology analysis, which used 2004 data (Wigley et al. 2007); (b)  stock assessments conducted during the 2005 Groundfish Assessment Review Meeting; and (c) various SAW/SARC analyses.

For most species, the VTR and Dealer databases provide similar values for the 2005 landings (Table 7).  VTR landings exceeded dealer landings in only six of the 39 species/species groups listed in Table 7.  Moreover, when two of the six species (offshore hake and red hake) are combined with white hake, the resulting VTR landings differ only slightly from the dealer data (1,996 mt vs. 2,063 mt, respectively).   For cases where the dealer landings exceeded the VTR landings (such as bluefish, scup, black sea bass, and monkfish), these discrepancies likely resulted from the inability to partition out the mandatory reporting landings (reflective of the VTR) from the state landings.  The differences for monkfish likely reflect misreporting of monkfish product forms (i.e., tails vs. whole fish) in the VTR database.

The results of the validation exercise show that for most species and species groups, the estimated landings derived using the NEFOP dataset do not differ significantly from the VTR values, with the 95% confidence interval of the estimated landings encompassing the VTR landings (Table 7 and Figure 1).  For three species (surfclams, ocean quahogs, and red crabs), the 95% confidence do not encompass the VTR or Dealer landings values.  However, there was no observer coverage of the 2005 fisheries for any of these species, and it is therefore not surprising that the estimated landings of these species do not approximate the VTR landings.  For the three hake species (red, white and offshore hake) and the two squid species (Illex and Loligo), there is some reporting of ‘mixed’ species such that the landings at the individual species level do not compare as favorably as at the combined (i.e., ‘mixed hakes’ or ‘mixed squid’) level (Table 7 and Figure 1).

The NEFOP, VTR, and DAS databases do not contain the requisite information to directly match trips (i.e., one-to-one match) across the three databases; hence, ad hoc methods were developed to accomplish matching.  Some misclassification of trips to various fishery sectors is therefore inevitable, and some of these misclassifications are evident in Table 2.   Two obvious examples of these misclassifications include: (a) VTR trips < NEFOP trips and (b) US/CAN area classification with MA area fished.  Some misclassification may also be due to the limited auditing of the VTR data resulting in overlapping trip dates, incorrect gear codes, and/or incorrect area fished.  With the NEFOP data, difficulties were sometimes encountered in identifying trips that ‘flipped’ between the B-day program and other programs.  In addition, when trips were matched between the VTR and DAS databases, 80 VTR trips had conflicting DAS codes (these were resolved by using the DAS code associated with the longest days absent; see Appendix B). When inconsistencies occurred between VTR gear and DAS access area, the VTR information was assumed to be correct.

Another constraint was the lack of master conversion tables in the NEFOP and VTR databases.  For the NEFOP, no master conversion factor table was available to convert dressed weight to live weight; hence, a conversion factor table developed for another analysis was used.  For the VTR data, a conversion between units of measure other than pounds (e.g. bushels, trays, bags, gallons, barrels) to pounds was needed. Again, a conversion factor table built for another analysis was thus used.[2]

In summary, a very broad stratification was used to encompass all species in the Northeast regional analysis.  Discard estimates provided in this report will differ from discard estimates developed separately in stock assessments because of differences in estimation methods and in spatial/temporal/fleet stratification schemes.

REFERENCES

Cochran WL. 1963.  Sampling Techniques.  J. Wiley and Sons.  New York.

Wigley SE, Rago PJ, Sosebee KA, Palka DL.  2007. The Analytic Component to the Standardized Bycatch Reporting Methodology Omnibus Amendment: Sampling Design, and Estimation of Precision and Accuracy (2nd Edition). Northeast Fish Sci Cent Ref Doc 07-09; 156 p.   http://nefsc.noaa.gov/publications/crd/crd0709/

ACKNOWLEDGMENTS

We wish to thank all the NEFOP observers for their diligent efforts to collect the discard information used in this analysis. 


Appendix A Table A1. Survival ratios for spiny dogfish and summer flounder, by fleet.


Appendix B. Method to assign DAS information to VTR trips

Overview

Matching trips between databases can be accomplished multiple ways.  A common way is to use exact matches between the vessel identifier and the sailing and/or landing dates (scenario 1 and 2).  This method work reasonably well when the trip endpoints are in agreement across databases.  When trip endpoints are not in agreement (e.g., Figure B1, scenario 3), a trip-midpoint matching process may improve the matching rate.  The trip-midpoint method matches trips by finding trips in database (A) where the midpoint of the trip falls between the sailing and landing dates of trips in the other database (B).  However, the trip-midpoint matching process is sensitive to which data set is used to define the start and end points of a trip and which data set’s trip midpoint is being bracketed (e.g., Figure B1, scenario 4 where the first VTR trip [A] would not be matched if the process uses the sailing/landing dates from the VTR [A] and the midpoint from the other database [B]).  One matching method that avoids this pitfall is to match trips that exhibit any degree of overlap.  The disadvantage of this approach is that it increases the number of multiple overlaps as seen in scenarios 4, 5 and 6.  The important questions to ask are: “can multiple matches be removed from the particular analysis?”, and/or, “are multiple matches likely to influence the results of the particular analysis?”  If the answer to these questions is “No.” then the overlap method is more likely to produce a larger matched data set compared to either the midpoint-matching process or the more traditional, exact matches (e.g., Figure B1, scenarios 1 and 2).

All matching processes will fail when trips that are true matches do not exhibit any overlap in the dates from the respective databases (e.g., Figure B1 scenario 7).  This situation is almost always caused by incorrect data entry of trip times in either of the two databases.  Because the VTR database contains self-reported data that is manually entered and only a limited amount of post-processing data auditing occurs, it is a reasonable assumption that the dates of VTR trips are less accurate than those of the other fisheries-dependent databases (e.g., Northeast Fisheries Observer Program [NEFOP], Days-At-Sea [DAS], Vessel Monitoring System [VMS], etc.).

VTR data conditioning

Examination of days absent (DA) from the VTR database revealed the presence of negative DA for approximately 1 % of the overall trips (1,227 of 123,766 trips) in 2005 (Figure B2).  All negative DA values are false. When negative DA values were less than -1.0 days, it was assumed that these were day trips with the times incorrectly entered.  To correct for this, all trips with DA <= 0 were assigned new start and end times of 00:00:01 and 11:59:59 (local times) on the start and end dates respectively.  Artificially increasing the duration of these trips in the VTR database resulted in a higher incidence of the situation observed in scenario 6 above.  Because these were generally day boats, taking a single trip per day, this was only an issue if a vessel had a negative DA trip and another fishing trip existed for the same day (i.e., multiple trips on the same day).  It should be noted that there were instances of multiple trips within the same day in the VTR data (1,038 of 123,766 trips) in 2005.  If any of these trips have negative DA, then this last assumption was violated, however the impact was small (37 trips out of 123,766 trips) in 2005.  This assumption would also have been violated if any of the negative DA trips had sailing dates that different from landing dates, however in 2005 there were no occurrences of this situation.

In addition to the concern that adjustment of the times of sailing and landing associated with negative DA trips would result in overlapping trips, there is also the possibility of overlapping trips in the rest of the trips (Figure B1, scenario 5).  No adjustment was made for these, but their presence is recognized.  The number of overlapping trips was less than 2.4 % of the total trips (2,910 of 123,766 trips) in 2005.

When matching two datasets for which optimization of the match rate is critical, it is important to have a reference match rate from another dataset to provide a point of reference.  For example, when matching DAS data to VTR data and only a 90 % match rate can be obtained, it may be that there is 10 % underreporting of VTRs such that a better match is not possible.  To provide a reference point for this analysis, the NEFOP data were examined.

Northeast Fisheries Observer Program (NEFOP) data conditioning

The NEFOP data identifies vessel using vessel hull number but not permit number.  Permit numbers had to be assigned to the NEFOP data to facilitate matches with other databases.  This was accomplished using the PERMIT database and matching on the sailing and landings dates.  An inability to match NEFOP hull numbers to the PERMIT database truncated the 2005 NEFOP data set[3] from 4,469 to 4,133.  Furthermore, all trips with DA <= 0 were deleted (a reduction to 4,118 trips for 2005 data; Figure B3).  An assumption was made that all remaining dates in the NEFOP dataset were valid and the match rate was assessed on the remaining trips (match rate among valid NEFOP trips).  There were 2 overlapping NEFOP trips in the 2005 data.

Days-At-Sea data conditioning

No data conditioning was performed on the DAS data set (Figure B4).

Match between VTR and NEFOP databases

In 2005, 3,642 of 4,118 NEFOP trips could be matched to a VTR trip (88.4 % match rate).  There were a total of 3,713 matched records.  Of the 3,713 matched records there were 8 VTR trips that matched multiple NEFOP trips and 70 NEFOP trips that matched multiple VTR trips (Figure B1, scenario 4).

Match between VTR and DAS databases

31,274 of 33,952 DAS trips could be matched to a VTR trip (92.1 % match rate).  There were a total of 32,088 matched records resulting in the assignment of DAS information to 31,362 trips. Of the 32,088 matched records there were 644 VTR trips that matched multiple DAS trips and 631 DAS trips that matched multiple VTR trips (Figure B1, scenario 4).

Based on the match results between VTR and NEFOP, the 92.1 % matching rate of DAS trips appears acceptable.  There are four likely reasons for the non-matching of the remaining 7.9 % of the DAS trips in the 2005 data:

  1. Under-reporting of VTRs (i.e., fishing occurred but no VTR was submitted/received for the trip);
  2. A VTR was not required for the trip (i.e., vessel was only setting gear or returned to port prior to engaging in fishing activity due to bad weather, mechanical breakdown or some other reason);
  3. A trip-stub exists in the DAS database that belongs to a longer DAS trip, but was not correctly assigned to a VTR trip because it falls outside of the sailing/landing dates reported on the VTR; and
  4. Due to incorrect reporting of trip dates to either database, a true match could not be determined when one exists (Figure B1, scenario 7).

For the purposes of this analysis, the critical issue was to correctly assign the appropriate DAS information (fishery code, DAS category code and access area) to the VTR trip.  So long as VTR trips were matched with the appropriate DAS information, it was unimportant that a DAS transaction could not be matched to a particular VTR trip (i.e., reason 3 given above).

It was important to ensure that the overlapping matches identified above (644 VTR trips matching multiple DAS trips and 631 DAS trips matching multiple VTR trips) did not result in conflicts with the assignment of DAS information to VTR trips.  This was determined by looking for VTR trips with multiple DAS code combinations (fishery_code||das_category||access_area).  In the 2005 data there were 80 VTR trips (< 0.3 % of total 31,362 assigned VTR trips) that were assigned conflicting VTR information resulting in 167 conflicting records requiring modification to reduce the conflict and assign a single DAS designation to these trips.  Based on a visual inspection of these 167 conflicting records a decision was been made to use the DAS designation with the longest days absent for a particular VTR trip.  If a tie was encountered in the days absent then the last DAS designation for a particular VTR trip was used.

www.nefsc.noaa.gov
NMFS Search
Link Disclaimer
webMASTER
Privacy Policy
(File Modified Jul. 01 2016)

This page has had 1 visits today, 1 visits this week, 75 visits this month, 801 visits this year