Reference Document Home | Publications Home
Results & Discussion
Appendix A
Appendix B.1
Appendix B.2

Northeast Fisheries Science Center Reference Document 08-17

Preparation of the Northeast Fisheries Observer Program Gillnet Data for Use in Bycatch Analyses of Protected Species

Melissa L. Warden1 and Christopher D. Orphanides2
1 Integrated Statistics, Inc. 16 Sumner Street, Woods Hole, MA 02543. Contact address: National Marine Fisheries Service, 166 Water Street, Woods Hole, MA 02543; 508-495-2084.
2 National Marine Fisheries Service, 28 Tarzwell Drive, Narragansett, RI 02543; 401-782-3283.

Web version posted September 2, 2008

Citation: Warden ML, Orphanides CD. 2008. Preparation of the Northeast Fisheries Observer Program gillnet data for use in bycatch analyses of protected species. US Dept Commer, Northeast Fish Sci Cent Ref Doc. 08-17; 44 p. Available from: National Marine Fisheries Service, 166 Water Street, Woods Hole, MA 02543-1026 or

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


Bycatch analyses of protected species (marine mammals, sea turtles, and seabirds) in domestic commercial fisheries off the Atlantic coast of the USA are usually conducted using data collected by the Northeast Fisheries Observer Program (NEFOP). In this paper, the preparation of a standard dataset of the 1989–2006 NEFOP gillnet data is described for use in bycatch estimation. Dressed landed weights were converted to live weights using established conversion factors, resulting in a 3.76% increase in observed weights in the NEFOP gillnet database. Data on additional variables such as bathymetry, sea surface temperature, chlorophyll a concentrations, oceanographic fronts, and distances from the coastline and bottom depth contours were obtained through Geographic Information Systems. Approximately 93% of missing depth and sea surface temperature values were replaced by GIS-acquired values. Other missing values were imputed through medians from representative strata. The standard dataset will reduce data preparation effort for bycatch assessments and enable consistency and reproducibility of bycatch estimates and other related analyses.


Gillnet fisheries are major components of the U.S. Northeast and Mid-Atlantic commercial fisheries. Incidental bycatch of seabirds, sea turtles, and marine mammals in gillnets are recorded by the Northeast Fisheries Observer Program (NEFOP). This is the only reliable source of fishery bycatch information for this region. The Protected Species Branch (PSB) of the Northeast Fisheries Science Center (NEFSC) regularly uses NEFOP data to estimate marine mammal bycatch (Belden 2007; Belden and Orphanides 2007; Belden et al. 2006; Bisack 2003; Rossman and Merrick 1999) and is also using the data to estimate seabird (Warden in prep) and sea turtle bycatch (Murray in prep).

The NEFSC Sea Sampling Observer Program, presently known as NEFOP, was initiated in 1989 to document the bycatch of fish, marine mammals, turtles, seabirds, and invertebrates taken during commercial fishing operations (Waring et al. 2007). Observer‑acquired information also includes fishery-related data such as landings and discards, as well as habitat, gear, and vessel characteristics (NEFSC 2008). Although the NEFOP database has a small percentage of missing values, the wealth of information in the database enables the derivation of additional variables useful in bycatch estimation.

This document outlines the augmentation of the NEFOP gillnet dataset for use in bycatch analysis. The creation of new variables from existing variables is described, as well as the addition of environmental data from outside sources and the imputation of missing values. The approaches described may also serve as a template for possible augmentations to other fisheries datasets.


2.1 Data

The NEFOP data represent observed fishing activity in state and federal waters from Maine to North Carolina (Figure 1). The highest concentrations of observed hauls are in the Gulf of Maine off of northern and central Massachusetts, and southeast of Cape Cod. The Mid-Atlantic gillnet fishery is primarily coastal with the highest densities of observed hauls off of Cape Hatteras and in the mouth of Chesapeake Bay.

From 1989 through 2006, NEFOP observed nearly 90,000 gillnet hauls (Table 1). Estimated observer coverage in the gillnet fishery (% by weight of total annual gillnet landings in observed hauls) varied annually from 1–7% in the Northeast region and 1–5% in the Mid-Atlantic (Table 2). During the 18-year period, observers recorded more than 4,500 incidental takes, including more than 2,700 seabirds, 900 cetaceans, 900 pinnipeds, and 70 sea turtles (Figure 2).

2.2 New Variable Creation

2.2.1 Geographical and regulatory variables

Regional closures and seasonal gear restrictions define important spatial and temporal features in the NEFOP gillnet data. Variables were added to the dataset to delineate the following: (1) regions within the study area; (2) harbor porpoise and large whale regulatory areas; and (3) the percentage of active marine mammal acoustic deterrent devices (i.e., pingers) on a net relative to the required amount.  Appendix A provides full descriptions of the new variables created.

2.2.2 Landings Converted from Dressed Weight to Live Weight

The NEFOP database is frequently used to estimate bycatch rates, which are subsequently applied to fishery effort/catch data to derive fishery-wide bycatch estimates for various species. Databases such as the Northeast Region Vessel Trip Report (VTR) database and the Commercial Fisheries Database System (CFDBS) are the principal sources for the detailed information on fishing effort and catch usually used in developing these estimates.

Landings are often used as a unit of effort in the gillnet fishery, as landings data are generally more complete and available than “true” effort data (i.e., time spent fishing). The landings can be whole or processed—defined in the NEFOP data as round or dressed, respectively. Dressed landings might be headed or gutted fish, mollusk meats, or parts such as monkfish tails or livers. In the VTR and CFDBS databases, dressed weights are converted to round weights (also known as live weights). For NEFOP landings to be directly comparable to landings in the VTR and CFDBS databases, dressed weights recorded by observers were also converted to live weights.

The conversion from dressed weight to live weight was based on three factors:

1) A common species code indicating the species and condition (i.e., round, gutted, tails, livers, etc.) of the landed product;
2) A NEFOP variable indicating whether the catch was round or dressed; [1]  and
3) The CFDBS conversion factor [2] for dressed to live weight for the species and product condition.

The product condition in the CFDBS table is called the grade. Table 3 lists CFDBS species codes, grades, and factors for converting dressed pounds to live pounds for major species affected by the conversion. The first three digits of the species code represent the species, and the final digit represents the grade. Grade indicators are not consistent across species (e.g., ‘0’ indicates tails for monkfish, but round weight for bluefish).

Grade indicators are not widely used as part of NEFOP species codes. For most species, NEFOP uses only the species code with the grade indicator for round catch (e.g., all landings of cod are recorded with a species code of 0818). The dressed/round variable must be relied upon to distinguish dressed weight from live weight. Major exceptions are monkfish and skates: NEFOP species codes for monkfish indicate whether the grade is round, tails, livers, cheeks, or belly flaps (note that gutted monkfish is not indicated—more on that type of conversion below); skate species codes indicate whether the grade is round or wings.

The monkfish and skate conversions from dressed to live weight were the simplest because their NEFOP species codes with grade indicators matched the species codes with grade indicators in the CFDBS table. For example, monkfish tails have the same species code (0120) in both databases and the conversion factor in the CFDBS table is 3.32 live pounds per dressed pound. Therefore, to derive a comparable weight between the databases, 100 lbs of monkfish tails in the NEFOP database was converted to 332 lbs (3.32*100) of live monkfish.

The grade portion of the species codes for skates sometimes contradicted the dressed/round variable. NEFOP is currently investigating the discrepancies, but for the 1989–2006 dataset used in the present undertaking, skate landings were converted only if the species code and the dressed/round variable agreed in indicating that the catch was landed as dressed (in this case, as wings).

Conversions for other species were more complex because the NEFOP species codes did not contain grade indicators. If catch was flagged as dressed according to the dressed/round variable, a grade was assigned from the CFDBS table using the smallest conversion factor for that species. To enable the conversion, the species code in the NEFOP data was changed to match the species code for the appropriate grade in the CFDBS table. For example, 100 lbs of dressed cod in the NEFOP data has the same species code as round cod (0818), but the dressed/round variable indicates that the catch is dressed. In Table 3, the grades for cod include both “dressed” and “gutted,” with conversion factors of 1.6 and 1.17, respectively. To avoid overinflating landings, the conversion from an unspecified form of dressed weight to a live weight would assume the catch to be gutted (as it has the smaller conversion factor). The species code for these 100 lbs would be changed to the CFDBS species code for gutted cod (0817) and the NEFOP landings would then be multiplied by 1.17 for a total of 117 lbs of live cod. This type of conversion was applied to monkfish, bluefish, cod, haddock, white hake, pollock, smooth dogfish, spiny dogfish, and silver hake (also known as whiting). A few other minor species were infrequently identified as dressed, but the catch amounts were so minimal over the time series that these were not converted.

After all dressed weights were converted to live weights, the live weight for all kept catch in each observed haul was summed, as was the live weight for all discarded catch in each haul. Catch recorded as kept but not landed in port [3] was treated as discarded, as the VTR and CFDBS databases contain only landed catch. Prior to 2001, all kept catch was listed with the same fish disposition so it was not possible to determine whether the catch was landed in port. For these years, all catch not discarded was assumed landed.

2.2.3 Environmental Data

A number of environmental variables from external sources were added to the NEFOP gillnet dataset for use in analyzing the effect of environmental factors on bycatch patterns. These included the following: sea surface temperature (SST), sea bottom temperature, indications of oceanographic fronts, surface chlorophyll a concentrations, bottom depth, bottom slope, distances from the coast and bottom depth contours, and North Atlantic Oscillation (NAO) values. The bottom depth and SST data from external sources were used to fill in missing values in the NEFOP bottom depth and SST variables (refer to section 2.3 for details).

Bottom depth, bottom slope, distances from the coast, and distances from bottom depth contours were added to the NEFOP database using electronic depth maps. Depth measurements were extracted from three topography datasets (in order of increasing spatial resolution): the NGDC ETOPOv2 [4] dataset; the Shuttle Radar Topography Mission [5] bathymetry dataset; and the National Geophysical Data Center (NGDC) Coastal Relief [6] dataset. Bottom slope in degrees of angle was calculated by applying the ArcGIS 9.2 Slope function to the ETOPOv2 dataset. Distance rasters for bottom depth contours and the coastline were generated using the ETOPOv2 dataset and the ArcGIS Euclidean Distance tool. Distances for fishing locations were retrieved from the rasters using the Extract Values to Points tool.   

Sea surface temperatures for NEFOP records were retrieved from satellite data. Five-day median SST image composites were created by merging SST from MODIS Aqua [7] , MODIS Terra [8] , GOES [9] , and AVHRR Pathfinder [10] satellite data sources. SST data were also retrieved from a NASA Jet Propulsion Laboratory climatology dataset. [11] Using ArcGIS 9.2 and IDL 6.3 software, SST was extracted from satellite images for each fishing location in the NEFOP database if the satellite data were available for that time and location. Point estimates for SST (derived from the value of one pixel, roughly 4 km x 4 km) were extracted from SST images, in addition to the median of the surrounding area (covering roughly 12 km x 12 km). For the final SST variable, point SST data were preferred over a 12 km x 12 km median temperature and five-day median images were preferred over climatology data. For a secondary SST variable, the best five-day median value was compared to the best climatology value and if the absolute difference was greater than 2.5˚C, the five-day median was deemed unreliable and the climatology was used for the final SST value. A third SST variable was created using the same process, but with a difference cutoff of 10˚C.

The strength and direction of oceanographic fronts were estimated from the five-day SST composites using Python scripts run in ArcGIS 9.2. The output frontal strength values were the maximum difference in temperature between a center SST pixel and its adjacent pixels, covering a roughly 12 km x 12 km area. The direction (aspect) of the frontal gradient was recorded as a compass value, 0°-359°, where 0° is due north.

Based on the TEMPEST routine (Mountain 1989), estimated climatological bottom water temperature was retrieved from the Marine Resources Monitoring, Assessment, and Prediction (MARMAP) dataset. The MARMAP data represent bottom temperature distributions generated from survey cruises conducted from 1977 to 1987 (Mountain and Holzwarth 1989). These bottom temperature measurements were also used in combination with SST data and bottom depth measurements to create new variables that document surface to bottom temperature differences and surface to bottom temperature change per meter of water depth, both of which reflect stratification of the water column.

Chlorophyll a image data were created using SeaWiFS and MODIS satellite data when available to create five-day composites for 1997–2005, and SeaWiFS high resolution data were used to create five-day composites for 2006. [12] A chlorophyll variable was created by sampling the chlorophyll at point locations and in a roughly 12 km x 12 km area surrounding the fishing locations. Point locations were assumed to be more representative of actual conditions and were used for the final chlorophyll value if available; if not, geometric mean values for the surrounding area were used. The final chlorophyll value was also log10 transformed for use in statistical analysis. [13]

Three types of North Atlantic Oscillation variables were created for the NEFOP dataset: the monthly mean NAO index, [14] three-month running averages of the monthly means, and a winter index. The winter NAO index averages the monthly means for December through March and is less noisy than the monthly index (Sandvik and Erikstad 2008).

Researchers have linked the NAO to variations in the abundance and distribution of the prey of marine mammals (GoMOOS, date unknown) and seabirds (Durant et al. 2004) via the influence of the NAO on surface water circulation patterns, but a time lag of several years may be evident.  To account for delayed effects, one-year and two-year lags were created for all three NAO variables.

Additional details and descriptions of the environmental variables are provided in Appendix A.

2.3 Missing Values

Missing values are inevitable in a dataset as extensive as the NEFOP database. However, for analysis of rare events such as bycatch, it is important to have as complete information as possible about each observed incidental take. In recent years, seabird, sea turtle, and marine mammal bycatch has often been estimated using a model-based approach (Murray 2004, 2005, 2006; Orphanides in prep.; Palka and Rossman 2001; Rossman in prep; Warden in prep.). When developing a model, a researcher may not have prior knowledge of which variables are most important in affecting bycatch rates, and may need to evaluate or test several potential variables. To allow the greatest flexibility in conducting such analyses on all documented bycatch events, an effort was made to replace missing values in the NEFOP database with the best imputed values. For continuous variables, the following hierarchy was used:

1) GIS-acquired data [15]
2) median from the same fishing trip
3) median from the same vessel and gear type in the same year and month
4) median from the same vessel and gear type in the same year
5) median from the same vessel in the same year only if the variability within the vessel/year stratum was low (defined as the coefficient of variation (CV) for the stratum equal to 0.10 or less, where CV = standard deviation divided by mean).

For categorical variables, the following hierarchy was used:

1) value from temporally nearest haul with the same gear type on the same trip
2) value from temporally nearest haul from the same vessel and gear type in the same year and month
3) value from temporally nearest haul from the same vessel and gear type in the same year.

Several steps were taken to ensure flexibility and transparency for data users. First, all variables were copied before being modified. Second, any modified variable was assigned an additional variable to indicate the source (i.e., observed, obtained through GIS, or imputed) of the value for each haul. For a handful of variables, atypical observations were corrected from the observer log or recalculated from available data and the source variable indicates these changes.

In addition to the source variable, a corresponding CV variable was created which reflects the coefficient of variation of the stratum from which a replacement median originated. This provides users with a way of detecting imputed values with high variability in the source. [16]

Appendixes B.1 and B.2 provide a list of variables that were processed for missing values, descriptions of these variables, and specifics on how certain missing variables were imputed.


3.1 Landings Converted from Dressed Weight to Live Weight

Converting dressed weights to live weights resulted in a 3.76% increase in total landed weights during 1994–2006. By species, the percent increase was as follows: skates 26%; smooth dogfish 6.4%; monkfish 5%; groundfish (cod, haddock, hake, pollock) 3.5%; bluefish 0.12%; and spiny dogfish 0.03%. The increase in monkfish weights was despite almost 30,000 pounds (0.5% of reported monkfish landings) of livers, cheeks, and belly flaps being removed because they do not contribute to the live weight (their conversion factors are 0).

The affected species were primarily in the Northeast, so the live weight conversion had little effect on landings in the Mid-Atlantic (4.5–5% increase in the Northeast and southern New England areas vs. 0.3% in the Mid-Atlantic area; see Figure 3 for regions). Annually, the greatest percentage increase in converted weights occurred in 2003 (8.6%) and 2004 (9.3%) and the least in 1995 (1.4%) and 1996 (1.7%) (Table 4).

Approximately 2,600 hauls (3% of all hauls) in the NEFOP gillnet database have no kept catch. About 150 of these hauls (0.17% of all hauls) resulted because beginning in 2001 kept catch not landed in port was considered to be discarded. These additional discards constituted approximately 77,000 pounds (0.55% of total landings during 2001–2006). Hauls with no catch pose complications when landings are used as a unit of effort in a bycatch model.  In these models, landings serve as the denominator for the bycatch rate and thus to avoid division by zero, hauls with no catch are generally either omitted from the analysis or their landings incremented from zero by a small amount.

Despite the conversions for dressed weights, total landings in live pounds should be considered a slight underrepresentation of the actual observed catches. This is because some skate landings that were likely wings were not adjusted due to the discrepancies between the NEFOP species code and the dressed/round indicator variable. The NEFSC is currently developing a comprehensive table for converting dressed weights to live weights in the NEFOP data, and this undertaking will address the discrepancies and minor species that were bypassed in the current dataset. This table will be applied to future NEFOP datasets and is an improvement over the current method.

3.2 Missing Values

Most variables processed for missing values (as listed in Appendix B.1) had less than 4% of their values missing, some less than 1%. Water temperature (wtmp) had the most missing values (25.5%) over the time series (1989–2006), but this variable was not recorded until May 1994. [17] During 1995–2006, 6.1% of the water temperature values were missing—still highest among the variables examined.

Acquisition of GIS-based variables was dependent upon nonmissing point locations as well as available data for the locations. Only about 2% of hauls were missing longitude and latitude information; in 14% of these cases, the median longitude/latitude values for the fishing trip were used to replace the missing data. SST and bathymetry values acquired through GIS were widely used in filling gaps in the observer-recorded water temperature and bottom depth: 93% of missing wtmp values were filled in by the GIS-acquired SST, while 92% of missing depths were replaced with GIS-acquired bathymetry data.  Summaries of missing and replacement values for all variables examined are provided in Appendix B.2.

For filling in missing values with representative medians, strata were designed to capture consistency in vessel fishing patterns. Low variability within a stratum indicates high consistency. For more than 50% of variables examined, the stratum with the lowest variability (as determined by the average CV across observations in that stratum) was vessel and year. This is because medians for this stratum were accepted only if the variability was low. On average, hauls from the same trip, and hauls from the same vessel, gear, year, and month show similar variability. For more than 60% of the variables, the stratum with the highest variability was the same vessel, gear, and year. Gear characteristics such as mesh size, number of vertical meshes on the net, and number of nets tend to show the least variability. Haul characteristics such as soak duration, haul duration, and bottom depth tend to exhibit the most variability. Appendix B.2 provides complete summaries of the variability across strata for all variables examined.

The imputation of missing values for the NEFOP gillnet dataset was undertaken to serve the needs of multiple users examining different outcomes (i.e., bycatch of different species). The approach taken was a type of conditional mean imputation, in which the means (or medians) from meaningful class levels of other observed covariates replaced missing values. Little and Rubin (2002) noted that conditional mean imputation can generate good point estimates of missing values, but such an approach ignores the uncertainty of the replacement values and may result in underestimation of standard errors, making inference invalid.  However, Ambler et al. (2007) found that conditional mean imputation performed well on simulated datasets with either 50% or 25% complete cases, and the logistic regression coefficients in their evaluation were unbiased and coverage for estimated confidence intervals was good. [18]  Nonetheless, because of the possible bias in standard errors, they caution against using p-values as criteria in model selection.  In a simulation study (Schafer and Schenker 2000) involving bivariate normal variables with missing values on one variable, conditionally mean-imputed data showed minimal differences from complete data in the width and coverage of confidence intervals at a missing value rate of 5%, but performance deteriorated when 10% or more of the data were missing. While neither simulation study is directly applicable to the NEFOP imputed dataset, the results provide a general sense that conditional mean imputation performs well and has minimal bias on estimates of variances when missing data problems are minor, as in the NEFOP dataset. However, information-based criteria (e.g., AIC) for model selection are recommended over hypothesis testing.

Even though conditional mean imputation is acceptable for minor missing data problems, methods of multiple imputation are generally recommended (Ambler et al. 2007; Little and Rubin 2002; Schafer 1997). In multiple imputation, several (typically 3-10) datasets are generated from an imputation model and analyses are performed on each dataset. Parameter estimates and standard errors are then pooled, incorporating the imputation variability. This method requires the outcome variable to be part of the modeling and so was not practical to implement for the current multiple-outcome dataset. Therefore, with the NEFOP imputed dataset, end users will need to decide between using the conditionally imputed medians and implementing an alternate imputation method.


In augmenting the NEFOP gillnet dataset, a consistent set of variables was provided.  The scope of habitat-related variables was broadened to help explain bycatch patterns, and NEFOP landings data were standardized to a measure congruent with other regional fisheries databases. A template was established for treating missing values, and imputed values were then substituted for missing data. The standard dataset will reduce data preparation effort for bycatch assessments and enable consistency and reproducibility of bycatch estimates and other related analyses.


Ambler G, Omar RZ, Royston P. 2007. A comparison of imputation techniques for handling missing predictor values in a risk model with a binary outcome. Stat Meth Med Res. 16:277–298.

Belden D. 2007. Estimates of cetacean and seal bycatch in the 2005 Northeast sink gillnet and Mid-Atlantic coastal gillnet fisheries. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 07-08; 16 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or

Belden D, Orphanides CD. 2007. Estimates of cetacean and pinniped bycatch in the 2006 Northeast sink gillnet and Mid-Atlantic coastal gillnet fisheries. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 07-20; 18 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02542-1026 or

Belden DL, Orphanides CD, Rossman MC, Palka DL. 2006. Estimates of cetacean and seal bycatch in the 2004 Northeast sink gillnet and Mid-Atlantic coastal gillnet fisheries. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 06-13; 24 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or

Bisack KD. 2003. Estimates of marine mammal bycatch in the Northeast (New England) multispecies sink gillnet fishery in 1996. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 03-18; 18 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or

Campbell, JW. 1995. The lognormal distribution as a model for bio-optical variability in the sea. J Geophys Res. 100(C7):13,237–13,254.

Durant JM, Stenseth NC, Anker-Nilssen T, Harris MP, Thompson PM, Wanless S. 2004. Marine birds and climate fluctuation in the North Atlantic. In: Stenseth NC et al., editors. Marine ecosystems and climate variation. Oxford (UK): Oxford University Press; p 95–105.

Gulf of Maine Ocean Observing System (GoMOOS).  Environmental prediction in the Gulf of Maine: forecast of North Atlantic right whale births [Internet]. [date unknown]. Portland (ME): GoMOOS; [cited March 5, 2008]. Available at:

Little RJA, Rubin B.  2002.  Statistical analysis with missing data. 2nd ed.  Hoboken (NJ): John Wiley & Sons, Inc; 408 p.

Mountain DG. 1989. TEMPEST: a computer program for estimating temperature on the Northeast continental shelf. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 89-02; 8 p. Available from: Information Services, NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026.

Mountain DG and Holzwarth TJ. 1989. Surface and bottom temperature distribution for the Northeast continental shelf. NOAA Tech Memo NMFS-F/NEC-73; 33 p. Available from: Information Services, NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026.

Murray KT. [in prep.] Sea turtle bycatch in U.S. Mid-Atlantic gillnet gear. NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026.

Murray KT. 2004. Magnitude and distribution of sea turtle bycatch in the sea scallop (Placopecten magellanicus) dredge fishery in two areas of the northwestern Atlantic Ocean, 2001-2002.  Fish Bull. 102:671-681.

Murray KT. 2005. Total bycatch estimate of loggerhead turtles (Caretta caretta) in the 2004 Atlantic sea scallop (Placopecten magellanicus) dredge fishery. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 05-12; 22 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or

Murray KT. 2006. Estimated average annual bycatch of loggerhead sea turtles (Caretta caretta) in U.S. Mid-Atlantic bottom otter trawl gear, 1996-2004. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 06-19; 26 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or

NEFSC (Northeast Fisheries Science Center), Fisheries Sampling Branch. 2008. Northeast fisheries observer program: Fisheries Observer Program Manual. Available from NOAA Fisheries, 166 Water St., Woods Hole, MA 02543-1026; retrieved March 5, 2008, from

Orphanides CD. [in prep.] Comparison of protected species bycatch estimating approaches: estimating harbor porpoise bycatch in U.S. Northwestern Atlantic federally managed gillnet fisheries. NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026.

Palka DL, Rossman MC. 2001. Bycatch estimates of coastal bottlenose dolphin (Tursiops truncates) in U.S. Mid-Atlantic gillnet fisheries for 1996 to 2000. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 01-15; 77 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or

Rossman MC, Merrick RL. 1999. Harbor porpoise bycatch in the Northeast multispecies sink gillnet fishery and the Mid-Atlantic coastal gillnet fishery in 1998 and during January-May 1999. Northeast Fish Sci Cent Ref Doc. 99-17; 36 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or

Rossman MC. [in prep.] Estimated bycatch of cetaceans in Northeast U.S. bottom trawl fishing gear. NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026.

Sandvik H, Erikstad KE. 2008. Seabird life histories and climatic fluctuations: a phylogenetic-comparative time series analysis of North Atlantic seabirds. Ecography. 31(1):73–83.

Schafer JL, Schenker N. 2000. Inference with imputed conditional means. J Am Stat Assoc. 95: 144-154.

Schafer JL. 1997. Analysis of incomplete multivariate data. London (UK): Chapman & Hall; 448 p.

Warden ML. [in prep.] Bycatch of common loons in Northeast and Mid-Atlantic gillnet fisheries in winter. NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026.

Waring GT, Josephson E, Fairfield-Walsh CP, Maze-Foley K, editors. 2007. U.S. Atlantic and Gulf of Mexico marine mammal stock assessments – 2007. NOAA Tech Memo NMFS NE 205; 415 p. Available from: NOAA Fisheries, 166 Water St, Woods Hole MA 02543-1026 or


[1] The dressed/round flag is not available before 1994 so the early data were not adjusted.

[2] The same factor used to adjust the VTR and CFDBS landings.

[3] This information is revealed by fish disposition categories: general kept; kept, transferred to another vessel; kept, used for bait; kept, eaten by the crew. Anything but general kept was treated as a discard.

[12] All chlorophyll data used in these images accessed through the NASA’s OceanColor website (

[13] Chlorophyll a concentrations span several orders of magnitude. A logarithmic scale results in a more manageable range. Log-transformed values generally follow a normal distribution ( Campbell 1995).

[14] Calculated by the Climate Prediction Center (see for method). This data was retrieved from the CPC website (, accessed 03/21/07).

[15] Bottom depth and sea surface temperature only

[16] If the median was derived from a single nonmissing value the CV will be missing because the standard deviation does not exist; the CV will also be missing if the mean used in calculating the CV is 0.

[17] Other variables not recorded until 1994 are noted in Appendix B.1.

[18] Coverage of the estimated confidence interval was calculated as the proportion of 1000 simulated datasets for which the 90% confidence interval included the true regression coefficient.
NMFS Search
Link Disclaimer
Privacy Policy
(File Modified Jul. 01 2016)

This page has had 4 visits today, 15 visits this week, 68 visits this month, 200 visits this year