NOAA/NMFS/NWFSC TM-31: Data Collection -- Groundfish: Chapter 2

CHAPTER 2
GROUNDFISH DATA COLLECTION IN WASHINGTON

Jack V. Tagart
Washington Department of Fish and Wildlife, Fish Management Program
Marine Resources Division, 600 Capitol Way North, Olympia, WA 98501, USA

Of the three U.S. Pacific coast states, Washington is unusual in being the home to a large fleet of distant-water vessels, some of which land a portion of their Bering Sea and Gulf of Alaska groundfish catches at Washington ports. Also, Washington is unusual in that it produces significant catches of groundfish from internal waters (Puget Sound). In this chapter, however, the descriptions of the groundfish fishery and data collection systems focus on the coastal fishery that operates outside of Puget Sound in waters south of the United States/Canada border.

Coastal groundfish are landed in Washington at more than 30 different ports. Major ports include Ilwaco (at the mouth of the Columbia River), Westport/Grays Harbor (in Grays Harbor), and Bellingham (in northern Puget Sound). In 1991 and 1992, these major ports accounted for 78% and 98% of the total groundfish landings in Washington (Tables 2.1- 2.2). Seasonally, the ports of Neah Bay (on the northwest tip of Washington at Cape Flattery) and La Push (at the mouth of the Quillyute River on the north central Washington coast) are also important, particularly for fixed-gear fisheries.

Principal fishing grounds for Washington-based trawlers extend from Tillamook Head located north of Tillamook Bay to Cape Flattery on the United States/Canada border (Fig. 2.1). Exceptions include the catcher-processor trawl fleet that fishes for Pacific hake (also commonly known as Pacific whiting) from northern California to Cape Flattery and occasional fishing by catcher-only trawl vessels as far south as Heceta Bank. Specific fishing grounds vary according to the size and fishing strategy of the vessel. For the trawl fleet, effort is concentrated around Cape Flattery Spit and the Juan de Fuca Canyon, Grays Canyon, and on the northern edge of Astoria Canyon (Figs. 2.2-2.3). Fishing occurs across depths from 10 to 500 fathoms but tends to be concentrated along the continental slope and lower portion of the continental shelf between 50 and 300 fathoms.

The Washington Department of Fish and Wildlife (WDFW) has no logbook program for non-trawl fishers and consequently, the distribution of non-trawl fishing effort cannot be described. However, it is well known to WDFW personnel that longline fishers, for example, work the continental slope off Cape Flattery Spit, including Nitinat Canyon, Juan de Fuca Canyon, and Astoria Canyon, while the jig fishery works the nearshore areas of the north Washington coast from Destruction Island to Cape Flattery.

Species landed vary by port and gear. In 1991 and 1992, principal species landed in the trawl fishery were rockfish, Pacific hake (also commonly known as Pacific whiting), arrowtooth flounder, and Dover sole (Tables 2.3-2.4). The primary species for fixed-gear fishers was sablefish.

The WDFW employs four full-time port samplers (scientific technicians) in three year-round positions and one half-year position. In addition, a senior scientific technician serves as immediate supervisor and coordinator of the regional samplers. The year-round staff are stationed at the Willapa Bay Field Station in Ocean Park (Ilwaco area), the Coastal Field Station in Montesano (Westport/Grays Harbor area), and the Field Office in Bellingham (Fig. 2.2). The seasonal port sampler works out of the North Coastal Field Station in Forks and monitors the ports of La Push and Neah Bay. The port sampling supervisor works out of the Coastal Field Station in Montesano.

All port samplers have a similar set of responsibilities, including collection and data entry of trawl logbook data (catch, effort, and area of catch), collection of biological data (length, weight, sex, age, etc.) for selected groundfish species, estimation of the species composition of rockfish landings, weekly collection and transmittal of preliminary groundfish landings, and communications and liaison with the commercial fishing fleet. The sampler stationed in Bellingham routinely samples landings in Bellingham, Blaine, and Anacortes. The sampler stationed in Montesano routinely samples Westport and Hoquiam, with additional sampling in Raymond, South Bend, and Bay Center. The Ocean Park sampler gathers data from vessels in the Columbia River ports of Ilwaco and Chinook. Seasonally, from April through September, the sampler stationed in Forks collects data from vessels in Neah Bay and La Push. Vessels operating in unsampled ports mail their trawl logbook data to Olympia or to one of the field offices. These data are then forwarded to a port sampler for data entry.

Washington's port samplers are involved with three principal sampling tasks: 1) collecting logbook data from groundfish otter-trawl fishers, 2) collecting biological information from designated groundfish species, and 3) collecting rockfish species-composition data from multispecies landings. Special sampling requests are also accommodated as time allows. The otter-trawl logbook system was initiated in 1953 to distribute the landed weight of groundfish reported on fish receiving tickets to the correct statistical catch reporting area and to estimate the effort expended to obtain the catch. The evolution of this trawl logbook system is reviewed in Kimura and Pattie (1982) and Clark (1986a, 1986b). The system has been identified by several acronyms over the years: in the 1960s it was RIRS (Randomized Information Retrieval System), in 1980 RIRS was replaced with TIS (Trawl Interview System), and in 1985 TIS became CTLS (Coastwide Trawl Logbook System).

Biological data have been collected using different sampling protocols through the years. As a rule, the sample has always been designated as a fixed number of fish as opposed to a fixed weight of fish. Sample size has varied from as many as 300 fish per sample to as few as 25 fish per sample. The sampling protocol has usually required drawing a simple random sample from the landed catch prior to sorting by the processor. In the past, some samples were obtained using a stratified sampling scheme for age structures (a fixed number of fish per cm interval) and simple random sampling for length data. Annually, the number of samples collected were reported in the minutes of the Technical Subcommittee of the Canada/United States Groundfish Committee. Current sampling strategies follow Kimura (1984): 25-100 specimens per sample and 10-25 samples per stratum (e.g., gear/area fished/time).

Rockfish species-composition sampling was initiated in the late 1960s. Prior to this time, reported landings from fish ticket receipts in the Pacific ocean perch market category, the principal rockfish species landed, were assumed to contain only Pacific ocean perch, i.e., they were thought to be uncontaminated by other species of rockfish, which were in limited demand and received little interest from fisheries managers. In the late 1960s, extensive foreign fishing dramatically reduced the Pacific ocean perch population, and domestic harvesters began to land increasing quantities of other deep-water "red" rockfish. Simultaneously, the market demand for other rockfish grew along with the interest of fisheries managers. Subsequently, the species-composition estimation program began. Since its inception, the program has employed a variety of estimation algorithms and sampling protocols. Tagart and Kimura (1982) report on estimation methods and sampling schemes used by the agency through 1984. In 1985, the estimation algorithms were changed, while the sampling methods remained largely the same. The estimation algorithm used through 1992 is reported in this document (see section 2.6). Updated algorithms for species-composition estimation from 1993 to date are under development.

Under the authority of the director of the Washington Department of Fish and Wildlife through published regulations in the Washington Administrative Code (WAC), fishers and processors are required to comply with the lawful requests of authorized employees. Under the provision of WAC 220-20-010(8): "It shall be unlawful for any person taking or possessing food fish or shellfish taken from any of the waters or beaches of the Columbia River, the state of Washington or the Pacific Ocean for any purpose to fail to submit such food fish or shellfish for inspection by authorized representatives of the department of fisheries [and wildlife]." Furthermore, under WAC 220-20-010(18), "It is unlawful for any person or corporation licensed by the department of fisheries [and wildlife] to fail to comply with the directions of authorized department personnel related to the collection of sampling data or material from food fish or shellfish. ..." Finally, WAC 220-44-080 makes it "... unlawful for any operator of otter trawl gear to fail to possess and maintain a Washington-Oregon-California Trawl Logbook while fishing in Coastal Marine Fish and Shellfish Management areas ..." Moreover, the vessel operator is required to keep the logbook aboard the vessel while fishing and to submit the completed logbook to department personnel immediately upon request.

Port samplers experience few problems with respect to obtaining biological or species-composition samples. The rapport between fishers, processors, and samplers is good, requiring few reminders to assist or comply with the request of a sampler. There are more frequent problems obtaining trawl logbook data, but these problems are generally associated with tardy logbook delivery. The WDFW has campaigned to get vessel operators to comply with the mandatory logbook system, requiring legible and complete logs delivered to the state in a timely manner. When an operator fails to deliver a log, the sampler mails a letter advising the operator that the log is overdue and should be delivered to their office in 10 days. Depending on the patience of the sampler and the history of the operator, the sampler may send a second letter requiring delivery of the delinquent log. Any subsequent reminders are made through the WDFW Enforcement Division. Lack of compliance can result in the issuance of a ticket and a fine. Overall, this process is effective, with compliance rates typically exceeding 85% (Table 2.5).

Washington's Marine Fish Receiving Ticket ("fish ticket") is an official document used to record the landed weight and value by species of designated marine food fish (Fig. A-1). Additionally, the fish ticket must identify the fisher, their address, vessel name, vessel registration number, the fishing gear, date of landing, area fished, the processor purchasing the fish, whether the fish were caught within three miles of the coast (state waters), within 200 miles (federal waters) or outside 200 miles (international waters), and the number and amount of fish retained by the crew for personal use. The Washington Department of Revenue levies a tax (2.14%) on the value of the landing.

Market categories recorded on the fish ticket are established by the director of the Department of Fish and Wildlife from the list of designated food fish (WAC 220-12-010) and with the assistance of fisheries managers in accordance with the selling practices of the fish processors. For example, sablefish are a designated food fish, but their landings are coded under multiple market categories depending on whether they are sold round or dressed and on their size and grade. Processors are provided a booklet by the WDFW called, "Catch Data Reporting Instructions to Dealers and Buyers," which contains the codes for various market categories. There are 18 preprinted codes and species descriptions on the fish ticket. Purchases of food fish in other market categories must be handwritten onto the fish ticket. By statute, the processor is required to list purchases of any species with an explicit trip limit in a separate market category, e.g., sablefish, Dover sole, and some species of rockfish. Rockfish purchases are currently recorded by the department in seven market categories: Pacific ocean perch, widow rockfish, yellowtail rockfish, canary rockfish, shortspine thornyhead, longspine thornyhead, and "other rockfish."

The WDFW issues, at no cost, sequentially numbered fish tickets to licensed buyers. Buyers are responsible for every ticket they are issued and must return either a completed or voided ticket in the number sequence in which they are used. Buyers are required to mail a completed fish ticket to the Department of Fish and Wildlife in Olympia within six days of receiving a landing.

In Olympia, marine fish tickets are separated from other tickets and previewed by a research analyst. The ticket is checked for completeness, assuring that all relevant information has been entered. The research analyst reviews species descriptions and market codes entered on the ticket and may replace an entered code with the appropriate code as required. For example, sablefish landings may be recorded by the buyer in multiple market categories by weight and product form (round or dressed). These categories are subsequently coded by the research analyst to one of the preassigned market categories. Missing data are retrieved by the research analyst by contacting the buyer and/or port sampler. The research analyst also reviews the catch reporting area, assuring that only one area is coded. When a buyer has noted catches from multiple catch reporting areas, the research analyst determines, through contact with the buyer or port sampler, where the bulk of the catch was taken and codes the ticket to that area.

After fish tickets have been previewed and precoded by the research analyst, they are keypunched and entered into the master database file. Keypunching typically occurs within a week of receipt of the fish ticket. After keypunching, the ticket data may still contain errors. Changes to the master database file are designated as new entries or updates. All new entries and updates are screened through an edit program that checks for missing data and erroneous codes, e.g., date, catch area, number of fish, pounds of fish, value, species codes, and days fished. Errors are reported in an error file that is reviewed by the research analyst. The errors are rectified by examining the original ticket and/or through further contact with the buyer or port sampler. Final corrections are again sent for keypunching and the master database file is updated. The error detection process is dynamic. Corrections to the master database file may occur at any time. Subtle errors undetected in the initial screening may be detected and corrected months later. At the end of a calendar year, the department prepares a preliminary "bound volume" of landed catch. The "bound volume" is a compilation of landed catch from commercial fisheries for all food fish and shellfish. This preliminary report is reviewed by research analysts and program managers for accuracy. Corrections are referred to data processing and a final "bound volume" is published two years after the landings were originally reported. At that time, the "bound volume" becomes the department's official record of landed catch and the fish ticket database for that year is closed.

For the coastal otter-trawl fishery, the department uses the standardized Washington-Oregon-California Trawl Logbook jointly developed by the Pacific Fishery Management Council and the states. There are no logbook requirements for other groundfish gears.

Otter-trawl vessel operators are required by law to maintain a complete and legible logbook and to deliver that logbook to an authorized representative of the Department of Fish and Wildlife on demand. Typically, logbooks are collected by the port sampler each time a fisher completes a trip. When the completed trip is unobserved by the sampler, the vessel operator is required to submit the logbook within 10 days of the termination of fishing or within 10 days of the end of the month, whichever comes first. In practice, the sampler may miss a vessel on a given trip and recover the log at the termination of a subsequent trip. When the sampler recognizes that an operator has completed a trip without submitting their log and the sampler has not been able to meet with the vessel operator to collect the log personally, a letter is sent to the operator reminding them of their obligation to submit the log. Ultimately, vessel operators with delinquent logs receive a warning or citation from the department's Enforcement Division as the case merits, accompanied by a demand to produce the log.

Logbook data are recorded to disk by the port sampler. Data are entered with the aid of a custom program that conducts cursory error checking. Coded logbook data are then sent to the Marine Resources Division in Olympia, where the data are copied and stored in separate files.

Raw logbook data are transmitted to the computer information consultant (CIC), a data specialist, in Olympia. Individual files of raw logbook data are each filtered through a comprehensive error checking program (Pattie 1987), at which point they become "processed trawl logbook data." Processed data are then aggregated into a single file and further processed into "tow-expanded logbook data" to account for tows that were not keypunched (Clark 1988a, 1988b). Tow-expanded data are subsequently processed with fish ticket data to generate the "expanded trawl logbook data."

The error screening program checks raw logbook data for "data out of range" errors in Loran coordinates, depth, fishing block, species, port, and trip type. It also screens for errors such as tow information entered in the wrong column and missing data. Records with errors are flagged in the database and a separate file is generated that describes the type of error and records the data line in the raw trawl data file where the error occurred. The CIC then has to rectify the error by reviewing the raw data or returning the coded data to the port sampler for clarification. More than 95% of the logs are free of coding errors after two passes through the error screening program.

Prior to 1993, WDFW obtained estimates of the species composition of rockfish landings almost exclusively by visual inspection of the landed catch. The procedure involved the observation of the top layer of one or more totes of landed fish. The frequency of occurrence of various species visible in the top layer was recorded by the sampler who subsequently converted the counts by number to weight based on average individual weights for various species. A standard set of mean species weights was maintained by the port sampling supervisor and used by each port sampler. The mean weights were derived from opportunistic basket samples of each species by dividing the cumulative basket weight by the number of fish in the sample. Samples from all ports were pooled and updated annually. The species composition was the proportion by weight of the species encountered. The visual method worked adequately provided the landing was dominated by no more than three species. When more than three species were present, samplers were encouraged to draw a weighed subsample of the landed catch, separate the species, and directly measure the weight of contributing species to the catch. There was no prescribed quantity for such a weighed sample. The method of determining the species composition was coded on the data entry form.

In 1993, sampling protocols were amended. Samplers were required to use a weighed subsample for any market category with more than three species in the landing. Furthermore, weighed subsamples were required for all landings in the "other rockfish" market category. Separate samples were taken for each of the five market categories (Pacific ocean perch, widow, yellowtail, thornyhead, and "other rockfish") used at that time.

Prior to 1994, samplers were encouraged to obtain as many species-composition samples as possible, but there was no prescribed number of samples. Landings could only be sampled when the rockfish catch was taken from a single Pacific States Marine Fisheries Commission (PSMFC) area. For example, when an "other rockfish" catch was taken from one PSMFC area and a widow or Pacific ocean perch catch from another, the landing could be sampled. However, if the catch of a single rockfish market category came from more than one PSMFC area, it could not be sampled.

In 1994, samplers were instructed to obtain 10-20 species-composition samples per 100 t of landed catch in each market category sampled. The PSMFC area continued to be the basis for qualifying whether a landing could be sampled.

The WDFW requires the processor to separate species regulated by trip limits from other species at the time they are unloaded. Samplers examine totes to verify that species have been separated appropriately. Their species-composition report will indicate the fish ticket reported weight of the landed catch and the observed species composition. Prior to 1993, all species-composition reports were coded to one of two market categories: nominal Pacific ocean perch or nominal "other rockfish." The term "nominal" implies that the market category, while ostensibly comprised of a single species, may actually be represented by additional species (e.g., Table 2.6). Since 1993, the number of market categories has increased to reflect the number of species regulated by trip limits.

Species-composition reports are coded onto forms by the port samplers and sent to a data specialist in Olympia. These data are keypunched and then screened for errors. A program checks for errors in area fished, port, and species codes; assures the correct year is coded and that the vessel code is valid; and verifies that the proportions by species sum to 100%.

The WDFW has collected rockfish species-composition reports for nominal Pacific ocean perch and "other rockfish" since the late 1960s (Tables 2.7-2.8). Since 1993, species-composition samples have been obtained for all market categories of species with designated trip limits. Species-composition sampling is conducted throughout the fishing year (Tables 2.9- 2.10), but is routinely undertaken only in the major ports.

In 1995, WDFW's species-composition sampling underwent a substantial revision. Species-composition samples were stratified by port/quarter rather than by PSMFC area/month. Single species market categories, which were those categories that had been previously demonstrated to contain primarily one species (at least 98%), no longer were sampled for species composition. These included Pacific ocean perch, widow rockfish, and yellowtail rockfish. Visual and weighed subsample techniques continued to be used for species-composition sampling for the remaining market categories: shortspine thornyhead, longspine thornyhead, canary rockfish, and "other rockfish." Revised catch estimation algorithms were adopted to reflect the change in sampling design.

The following example illustrates some of the protocols and decisions involved in sampling commercial groundfish landings in Washington prior to 1995, when field techniques involved in the species-composition sampling program were modified (see section 2.4). A vessel makes a landing that includes 4,000 lb of nominal yellowtail rockfish, 3,500 lb of nominal widow rockfish, 800 lb of nominal Pacific ocean perch, and 7,000 lb of "other rockfish." The sampler observes the unloaded catch in 18 totes awaiting processing. To determine whether the catch can be sampled for rockfish species composition, the sampler reviews the vessel's logbook, which indicates that all the yellowtail rockfish and Pacific ocean perch were taken in PSMFC area 3CS, all the "other rockfish" were harvested in area 3B, and the widow rockfish were caught in both areas. The totes containing widow rockfish cannot be sampled for species composition because the catches were made in multiple PSMFC areas, but the remaining totes can be sampled.

Yellowtail rockfish and Pacific ocean perch are species regulated by trip limits. The processor purchasing the fish has segregated these species from the "other rockfish"; however, a small amount of "other rockfish" species remains with the Pacific ocean perch. The sampler finds only one tote with Pacific ocean perch and notices that it is contaminated with only redstripe rockfish. Using a visual estimation technique, the sampler counts and identifies all of the fish that can be observed on the top layer of the tote. There were 41 Pacific ocean perch and 3 redstripe rockfish. Using standard mean weights, 1.54 lb for Pacific ocean perch and 1.29 lb for redstripe rockfish, the sampler records a species composition of 94% Pacific ocean perch and 6% redstripe rockfish. The sampler observes that the yellowtail rockfish have been successfully separated from the other species, i.e., no contamination, and reports a species composition of 100% yellowtail rockfish.

The "other rockfish" are contained in eight totes. Visual inspection of the top layer of these totes indicates that there are more than three rockfish species present in the landing. The sampler determines that there is adequate time to obtain a weighed subsample of the catch. Selecting one of the eight totes, the sampler sorts the catch into baskets containing single species. Each basket is weighed on a platform scale with a composition as follows: 400 lb of bocaccio, 200 lb of redbanded rockfish, 100 lb of splitnose rockfish, 75 lb of sharpchin rockfish, 50 lb of Pacific ocean perch, 30 lb of rosethorn rockfish, and 20 lb of greenstriped rockfish. The species composition is then 46% bocaccio, 23% redbanded rockfish, 11% splitnose rockfish, 9% sharpchin rockfish, 6% Pacific ocean perch, 3% rosethorn rockfish, and 2% greenstriped rockfish.

The WDFW routinely collects biological samples from Dover sole, English sole, arrowtooth flounder, Pacific cod, lingcod, yellowtail rockfish, widow rockfish, canary rockfish, black rockfish, and sablefish. Other species are sometimes sampled in response to specific requests. Samples are stratified by species, gear, and area fished.

The sampling level is specified by a research scientist in consultation with other staff. Annually, the table of requested samples is given to the port sampler's supervising senior technician who subsequently distributes the list to all samplers and coordinates the data collection activities. The port sampling supervisor compiles semi-monthly summaries of samples collected at each port and shares this information with all samplers. Samplers are requested to collect a specified number of samples per species for combinations of gear/area fished/month. The supervisor assures that the samples are collected on schedule and reminds port samplers of missing samples when necessary.

Port samplers temporarily store biological samples (age structures) in their field office and periodically transfer them to Olympia for processing and permanent storage. In Olympia, the samples are delivered to the technician responsible for age determination. The technician logs the samples and stores the structures as required (spines and interopercles to a freezer, and otoliths to a storage chest). The accompanying data forms are sent to a CIC who enters the data into the WDFW's Biological Data System (BDS) (Kimura and Cross 1983). Once ages have been determined, the age reader enters the information into the BDS database.

New or recently hired samplers, as a part of their initial job orientation, receive training on how, where, and when to sample as well as how to complete the many data entry forms. All groundfish biological data are recorded in the same format. Paper records of sablefish samples are also coded and maintained in the format requested by the National Marine Fisheries Service.

Vessels arriving at a port are selected for sampling opportunistically, i.e., non-randomly. Moreover, only those vessels whose catch can be identified as coming from a prescribed gear/area fished stratum are eligible for sampling. Individual specimens are typically drawn at "random" by taking fish from one corner (top to bottom) of an unsorted tote of landed fish. Occasionally, the sampler will obtain specimens from the fillet line prior to processing and/or from discarded carcasses (frames) after processing. As mentioned earlier, WDFW's sampling protocol, with the exception of sablefish, requires drawing a fixed number of fish for each sample (see section 2.1.2).

During data entry, the CIC checks data manually for obvious omissions and gross errors. There is no formal error checking program to scan the biological data; consequently, some errors are undetected until the data are analyzed. Types of errors that may occur include transposed size or age data (trapped when data are scanned for outliers) and missing integers on individual specimen weight data, which occasionally occur because samplers neglect to enter the integer portion of the weight when recording data in the field.

The BDS is currently a 20-megabyte computer database containing more than 1 million records. Data records date back to 1954 for some species (Table 2.11).

The following sections describe estimation algorithms for the determination of groundfish catch and effort and rockfish catch by individual species. These algorithms form the basis of a catch reporting system that has undergone numerous changes over the years. The algorithms used to derive point estimates of catch and effort are applicable for all years of data collection by WDFW unless otherwise noted; however, the estimates of variance are applicable for specified periods only.

In Washington, fish ticket receipts report the actual weight of landed fish by date, port, management area (a unique WDFW defined area), and market category (species or species group). However, area information on these receipts is imprecise because it is limited to only one reporting area when in fact one or more areas may have been fished. For coastal fisheries, the greatest volume of landings are taken by the trawl fishery, and each trawler is required to maintain a logbook with tow-by-tow records of their fishing location, towing time, depth fished, and estimated (hailed) weight of species landed (retained catch only). The fisher's recorded tow-by-tow catches cannot be used directly to describe landed weight because the logged weights are estimated by the skipper with varying degrees of accuracy, minor catches of some species are never recorded by the skipper, and the tow sampling rate (see section 2.6.1, tow-expanded values) used by the department has been less than 100% in some years (Clark 1988b).

Logbook data do represent the best information available on the relative distribution of catch by statistical catch reporting area (10 minute latitude/longitude block, WDFW statistical area, PSMFC and/or International North Pacific Fisheries Commission (INPFC) area). In addition, logbook data are the only source of information on trawl effort (hours fished), depth fished, and net type used. In order to provide tow-by-tow data that correspond with fish ticket totals, ticket landings are "distributed" over the logbook data in proportion to the tow-expanded values (Fig. 2.4). Tow-expanded values are the hailed weights multiplied by the ratio of total to sampled tows for a trip. Fish ticket receipts for a given vessel trip are not matched with the vessel's logbook entries. Rather, the expanded tow-by-tow logbook weights are multiplied by the ratio of fish ticket to expanded logbook weight. The distribution is performed within each of many primary strata defined by year, trip type (e.g., coastal, Puget Sound), month, port cluster (a group of ports), and species (Clark 1988b).

To the extent possible and practicable, the notation below utilizes unique variable and index values. The defined values are applicable for all subsequent equations unless specifically redefined. In addition, "dot" notation is sometimes used to represent summation over a specific index, e.g.,

In addition to multiple catch variables, there are multiple species indices. Fish tickets and logbooks report catches in market categories. A market category may represent either single or multiple species, e.g., Pacific cod, "other flatfish," or "other rockfish." Rockfish species-composition reports record catches in single and multiple species groups as well, but the groups may be different than those represented by the market category. For example, a landing recorded on a fish ticket receipt in the "other rockfish" market category could have a species composition as follows: 95% silvergrey rockfish, 3% redbanded rockfish, and 2% "other rockfish." In this case, the "other rockfish" in the species-composition report is a different collection of species than the "other rockfish" market category reported on the fish ticket. Consequently, one index is needed to represent the market categories, and a second index to represent the species or species groups within the market categories.

Finally, a few remarks are needed on the relationship between trips, fish ticket receipts, and species-composition samples. Logbook data are used to count the number of trips within a primary strata. For any given trip, a fisher may sell his catch to one or more buyers, each of whom will complete a fish ticket sales receipt. Therefore, the number of fish tickets is greater than or equal to the number of trips. For each trip fished and each landing at a unique buyer, a sampler may obtain one or more rockfish species-composition samples. Prior to 1995, multiple species-composition samples could be taken to reflect catches from different statistical catch reporting areas or to report the species composition from multiple market categories. Since 1995, multiple species-composition samples are taken only when multiple market categories from the same landing are sampled.

The total weight recorded on a fish ticket is the landed weight sold to a specific buyer. Some fishers sell the retained catch from a single trip to multiple buyers; therefore, the landed weight reported on a single sales receipt may not be the trip weight of a species. Some samples may be obtained from a portion of the landed weight at a single buyer. This is sometimes the case when sampling rockfish landings for species composition. So, the total weight of the sampled rockfish catch may not be the total weight reported on the fish receiving ticket nor the total weight of the trip.

The WDFW has recorded tow-by-tow logbook data since 1985. The logbook system was designed to accommodate subsamples of tows. Originally, approximately every fourth tow was keypunched from each logged trip; currently, all of the recorded tow-by-tow data are entered into the logbook database. When tow-by-tow data are subsampled, hailed weights of the subsampled tows have to be expanded to represent the weights of all tows in the trip. This process produces the "tow-expanded" values, which are simple random sampling estimates.

Fish ticket landings are "distributed" over the tow-by-tow logbook data in proportion to the tow-expanded values. In doing so, the ratio between logged effort and logged catch is preserved in the distributed catch and effort. The estimates of distributed catch and effort must accommodate several special cases. To preserve the absolute sum of landed catches recorded on the fish tickets, adjustments are necessary for catches of species that are reported on the fish ticket but not in the logbook data. In addition, effort is missing from some logbook tows. To preserve the catch-per-unit-effort (CPUE) ratio represented by the sampled logbook catch, missing effort is estimated from the ratio of distributed catch for each tow to CPUE for all tows.

The distributed fish ticket values (catch weight and effort) are estimated within each of many primary strata determined by trip type (g), year (y), month (z), and port cluster (p). Ports are clustered to minimize the number of cells with missing values (i.e., primary strata containing only logbook or only ticket data) (Table 2.12). There are five trip types: Alaska, Canada, outside, inside, and deep inside. Outside trips represent catches from the Washington, Oregon, and California coast beyond the Strait of Juan de Fuca. Inside trips represent catches from Puget Sound and the Strait of Juan de Fuca. Deep inside refers to the inner reaches of Puget Sound. The catches from inside and deep inside trips are pooled for the purposes of distributing catch to avoid too many empty cells. However, the trip type is retained with the distributed values in the final catch database.

Then the distributed catch corresponding to each record of tow-expanded catch is,

There are three cases to be considered when computing distributed effort: 1) tow-expanded effort is unknown, (i.e., the duration of the tow was never recorded in the logbook or the catch of this species appeared on a fish ticket but was never recorded in the logbook, which makes this a "TICKETS" data line for which there is no corresponding effort information), 2) tow-expanded effort is known, but the distributed catch is less than the tow-expanded catch, and 3) tow-expanded effort is known, but the distributed catch is greater than the tow-expanded catch.

For clarity and convenience in the following exposition, a variable is defined to represent expanded trawl effort. However, from a data-processing perspective, effort is just another "species." Recall from equations 2.2-2.3 that t_ij represents the tow-expanded catch of species i and that effort is given a dummy species code (599).

The proportion of each rockfish species in a stratum is determined from a simple ratio estimator following Cochran (1977). Estimated landed weight within a year (y), month (z), and PSMFC area (a) stratum is the sum over all market categories of the product of the estimated weighted proportion by species and the distributed fish ticket catch by rockfish market category.

Prior to 1985, submission of trawl logbooks was voluntary. Tow-by-tow data reported in submitted logs were aggregated based on area of catch prior to coding. Since 1986, trawl logbooks have been mandatory, and although subsamples of logbook tow-by-tow data were originally keypunched, the WDFW has subsequently recorded all of the tow-by-tow data obtained from trawl logbooks. Consequently, components of variance due to subsampling tows have been eliminated. Moreover, because we are quite successful at recovering logbooks for nearly all trips (> 85%), the sampling system is, in effect, a complete census and all variance components go to zero, with the exception of variance due to rockfish species-composition sampling. The variance formulae presented below are applicable when tows are subsampled and/or when total trips are subsampled. They are provided as a record of the algorithms used in the CTLS and as an aide for those who may be considering a logbook system that relies on subsamples of tows and trips. Tow-expanded values

Variance for the tow-expanded catch is derived from standard methods. The variance for t_i is calculated as,

Complications in calculating variances of distributed quantities are discussed in detail by Clark (1988b). Complications arise as a consequence of summing a subset of distributed values from each of several primary strata into a new aggregate of interest. For example, suppose one wishes to estimate the annual catch and variance for English sole caught in PSMFC area 3B. Catches will have to aggregated across primary strata for each month and across strata representing different port clusters. The variance of the total catch of English sole is a composite of the variance due to sampling tows and trips and the variance of the estimated proportion of the fish ticket catch represented by the logged catch. The variances are computed within each of the contributing primary strata and then summed across strata to represent the quantity of interest.

The variance for the rockfish proportion by species (the ratio estimate) is computed following Cochran (1977).

Let m be the number of species-composition sampled landings within a market category.

Then the variance of the estimated proportion of rockfish species o in market category i is,

Reported landings of groundfish by non-trawl gears are taken directly from fish ticket receipts. Area of catch is converted to a PSMFC area based on the reported WDFW management area. Effort is not estimated for non-trawl gears.

Catch-at-age algorithms are variations on a theme. One has to compute the proportion of the total catch weight attributable to each age class, and divide that weight by the mean weight-at-age to compute the number caught-at-age. Estimates of catch-at-age can be derived through weighted or unweighted methods. Unweighted methods pool data from all samples within a stratum, giving each sample equal weight. Since individual fish are only weighed periodically, the weight of each specimen is estimated from an allometric length-weight relationship. For the pooled samples, the frequency of occurrence at each age is determined, along with the total weight-at-age. The proportion-at-age, based on weight, is then calculated. These proportions are multiplied by the total catch weight and then divided by the mean weight-at-age to generate catch-at-age in numbers of fish. Weighted methods for estimating catch-at-age weight each sample prior to aggregating the observations within a stratum. The sample is typically weighted by the weight of the total catch from which the sample was drawn. Other weighting methods convert the fixed number of fish sampled into a sample of fixed weight (Tagart 1991). Variance measures for the estimates of catch-at-age have not been developed.

Under the "redefined PacFIN" protocols (see chapter 6), the WDFW submits the complete, detailed, line-by-line fish ticket data to PacFIN. Also submitted is a matrix of proportions that assign the fish ticket data to PSMFC area according to adjustments based on trawl logbook data as described above. Finally, the WDFW provides an additional matrix of proportions that assign the rockfish species composition to the area-adjusted catches of each rockfish market category reported in the fish ticket data. Variances in the form of coefficients of variation are also presented for the area-adjusted catch and species-composition data.

The point estimates of trawl catch will be reported in PacFIN by PSMFC area and port and should be precise replicas of WDFW final data. The variances only apply to the area stratification and would not be applicable for catch estimates by port or any other aggregate.

Under the original PacFIN protocols, the WDFW submitted aggregated fish ticket data rather than individual ticket data. Based on the statistical catch reporting area coded on the fish ticket, catch data were distributed by INPFC area rather than PSMFC area. There was no incorporation of catch area adjustments from logbook data. Rockfish market categories on fish tickets were aggregated into two groups: nominal Pacific ocean perch and nominal "other rockfish." The nominal rockfish catch data were distributed to species based on a matrix of proportions from our rockfish species-composition data; however, the species-composition data were stratified by INPFC area rather than PSMFC area. So, the original PacFIN rockfish catch data accurately represented the total statewide catch by port and market category, but were often a poor representation of the catch by area and species. It should be noted that the original PacFIN data were expected to be replaced by logbook adjusted data as soon as the appropriate data became available; unfortunately, that turned out to be nearly 14 years later.

This chapter was prepared with the assistance of Martha Rickey, Farron Wallace, Sandra Rosenfield, Han Lin Lai, Brad Pattie, Cathy Evans, and Mel Stanley. Their contributions are gratefully acknowledged.

Clark, W. G. 1986a. Standardization of Washington's historical trawl logbook data. Wash. Dep. Fish. Prog. Rep. No. 229, 16 p. plus appendices.

Clark, W. G. 1986b. Washington's trawl logbook data, past and present. Wash. Dep. Fish. Prog. Rep. No. 228, 62 p.

Clark, W. G. 1988a. Operation of the coastwide trawl logbook system. Wash. Dep. Fish. Manuscr. Rep., 23 p.

Clark, W. G. 1988b. User's guide to Washington's expanded trawl logbook data. Wash. Dep. Fish. Prog. Rep. No. 271, 21 p.

Cochran, W. G. 1977. Sampling techniques, 3rd edition. John Wiley & Sons, New York, 428 p.

Kimura, D. K. 1984. Determining appropriate sampling levels for estimating age distribution from trawl landings. Wash. Dep. Fish. Tech. Rep. No. 80, 18 p.

Kimura, D. K., and C. M. Cross. 1983. Biological Data System (BDS) reference manual. Wash. Dep. Fish. Prog. Rep. No. 191, 26 p. plus appendices.

Kimura, D. K., and B. H. Pattie. 1982. Trawl Interview System (TIS) reference manual (beginning 1980). Wash. Dep. Fish. Prog. Rep. No. 159, 29 p. plus appendices.

Pattie, B. H. 1987. The Coastwide Trawl Logbook System (CTLS) error scan program (CTLPRO), and some common errors detected during 1986. Wash. Dep. Fish. Brief. Rep. MF87-002, 21 p. plus appendices.

Sen, A. R. 1986. Methodological problems in sampling commercial rockfish landings. Fish. Bull., U.S. 84:409-421.

Tagart, J. V. 1991. Population dynamics of yellowtail rockfish (Sebastes flavidus) stocks in the northern California to southwest Vancouver Island region. Ph.D. Thesis, Univ. Washington, Seattle, WA, 323 p.

Tagart, J. V., and D. K. Kimura. 1982. Review of Washington's coastal trawl rockfish fisheries. Wash. Dep. Fish. Tech. Rep. No. 68, 66 p.

let	m	be the number of tows recorded (i.e., sampled),
	M	be the total number of tows logged (i.e., fished),
	x_ij	be either the total hours fished, or the hailed catch of a single species (e.g., sablefish) or species group i (e.g., "other rockfish") from tow j, and
	t_ij	be the tow-expanded catch (or effort).

let	c_i	be the fish ticket catch of species i, and
	d_ij	be the distributed ticket catch of species i on tow j.

let	e_j	be the tow-expanded logbook effort on tow j,
	e_j*	be the distributed effort, and
	u_gh	be the catch per unit effort for all reported species (total tow-expanded logbook pounds per hour fished) for trip type g and net type h (bottom trawl, roller trawl, mid-water trawl, any trawl) computed from all sampled tows where effort was recorded.

let	x_q	be the sum of all logbook values within a primary stratum that qualify for a subset of interest, and
	x.	be the sum of all logbook values in the stratum.

let	t_q	be the sum of all qualifying tow-expanded logbook values (i.e., the sample catch or effort logbook data for the specified stratum of interest),
	t.	be the sum of all tow-expanded logbook values in the specified stratum of interest,
	n.	be the number of tows contributing to t.,
	d	be the sum of distributed fish ticket values in the primary stratum, and
	fc	be the relative completeness of logbook coverage in the stratum.

CHAPTER 2 GROUNDFISH DATA COLLECTION IN WASHINGTON

CHAPTER 2
GROUNDFISH DATA COLLECTION IN WASHINGTON