Characterisation and CPUE analyses of the fishery in FLA 3

M. P. Beentjes 1 M. J. Manning 2

1NIWA P O Box 6414 Dunedin 9059

2NIWA Private Bag 14901 Wellington 6241

New Zealand Fisheries Assessment Report 2010/27 September 2010

Published by Ministry of Fisheries Wellington 2010

ISSN 1175-1584 (print) ISSN 1179-5352 (online)

© Ministry of Fisheries 2010

Beentjes, M.P.; Manning, M.J. (2010). Characterisation and CPUE analyses of the flatfish fishery in FLA 3. New Zealand Fisheries Assessment Report 2010/27.

This series continues the informal New Zealand Fisheries Assessment Research Document series which ceased at the end of 1999. EXECUTIVE SUMMARY

Beentjes, M.P.; Manning, M.J. (2010). Characterisation and CPUE analyses of the flatfish fishery in FLA 3.

New Zealand Fisheries Assessment Report 2010/27 .

The objective of this project was to characterise the FLA 3 fishery and analyse CPUE trends in the fishery. In addition, we also investigated ancillary data that included processor’s landings data, SeaFic logbook programme, and east coast South Island (ECSI) Kaharoa trawl surveys.

The fishery characterisation data were extracted from the Ministry of Fisheries catch effort database Warehou from Catch Effort Landing Returns (CELRs) and Trawl Catch Effort Landing Returns (TCEPRs) for 1989–90 to 2006–07, and Net Catch Effort Landing Returns (NCELRs) for 2006–07, for statistical areas within FLA 3. The key features of the aggregated flatfish species fishery in FLA 3 include: 1) no strong seasonal trend in catch; 2) important statistical areas are those contiguous with the coast (020, 022, 024, and 026); 3) virtually all catch outside Te Waihora is taken by bottom trawl, and setnet within Te Waihora; 4) most of the flatfish catch was caught by vessels recording FLA (generic flatfish species code) as the target species with modest catches with target red cod (RCO), and lesser catches with target New Zealand (ESO) and lemon sole (LSO); 5) virtually all flatfish catch is landed from vessels completing CELRs (up to 2006–07); 6) the only three flatfish species caught in any quantity by trawling are ESO, LSO, and to a lesser extent sand (SFL). Catches of these species are spread throughout the coastal statistical areas from 020 to 030, although LSO catch was greater in 026; 7) the setnet catch is virtually all from statistical area 022 (assumed to be Te Waihora), and the only species caught in any amount are black flounder (BFL), SFL, and yellowbelly flounder (YBF).

Kaharoa ECSI trawl surveys indicate that LSO is the main flatfish species followed by ESO and SFL in winter, and ESO and greenback flounder (GFL) in summer. SFL inhabit the shallow coastal areas down to and probably inside 10 m, ESO predominantly in 10 to 30 m, with LSO extending deeper again into the mid, and occasionally, outer continental shelf. Although the Kaharoa trawl net and survey design are not optimised for flatfish species, the relative proportions of species may be reasonably well represented within the depth ranges surveyed. The Sanford’s data and logbook scheme flatfish species mix are consistent with this finding. The flatfish species length frequency distributions from the Kaharoa and industry logbook scheme are similar and there are some indications of modes that represent age classes. There are insufficient data to plot length frequencies by survey to look at recruitment strength.

An algorithm was applied to the aggregated flatfish groomed and merged (rolled-up) dataset to estimate the proportion of each flatfish species in the generic coded FLA catch. Allocation was based on 1) the associated non-generic estimated catches for the associated fishing records or trip and 2) Sanford’s Ltd Timaru dataset of species-specific landing weight data collected from 1992–93 to 2000–01 and 2003–04 to 2007–08 (there were no data for 2001–02 and 2002–03). Standardised CPUE indices were obtained by fitting Generalised Linear Models to the imputed (generic FLA codes converted to species specific codes) dataset. Separate CPUE analyses were carried out, for the bottom trawl fishery (FLA, ESO, LSO, and SFL) and the Te Waihora setnet fishery (FLA, BFL, SFL, YBF). Trawl fishery CPUE analyses for all species combined and the three key species, ESO, LSO, and SFL show generally similar trends of increasing CPUE in the mid to late 1990s, followed by a steep decrease, after which ESO has a secondary peak about 2001–02 followed by a steep decline, LSO increases again after about 2003–04, and SFL fluctuates with no clear trend. The trawl CPUE analyses are characterised by large fluctuations from year to year, reflecting few year classes and a fishery that is highly recruitment driven. The CPUE trends for all three species are consistent with biomass trends from Kaharoa surveys, commercial landings in FLA 3, and processor’s catch data, providing evidence that the CPUE indices generated for these three species are a proxy for abundance. The setnet CPUE models are marked by large fluctuations in the annual CPUE indices with no clear trends, and wide confidence limits around the indices, the latter a reflection of having only about 7 participants in the fishery. Unlike the trawl CPUE indices, there are no proxies of abundance to compare with the setnet CPUE indices to test for plausibility.

3 1. INTRODUCTION

The flatfish commercial fishery in New Zealand comprises eight species from four genera: four (black flounder, retiaria ; greenback flounder, R. tapirina ; sand flounder, R. plebeia ; yellowbelly flounder, R. leporina ), two soles (lemon sole, Pelotretis flavilatus ; New Zealand sole, Peltorhamphus novaezeelandiae ), plus brill ( Colistium guntheri ) and , ( Colistium nudipinnis ) (Kirk 1989, Colman 1994, Ministry of Fisheries 2006). Witch ( Arnoglossus scapha ) is also caught, but is less desirable and seldom landed. The Management practise of combining flatfish reported landings from all eight species under the code FLA has made it difficult to monitor abundance or catch trends of individual species. It is understood that the proportions of each species vary within the reported catch among the four flatfish QMAs (FLA 1, FLA 2, FLA 3, and FLA 7), but the relative annual contribution of each species is unknown since there are no landings for individual flatfish species documented (Ministry of Fisheries 2008a). In status of the stocks, the 2008 plenary report states that because of highly variable recruitment and few year classes …”a constant catch at the level of the current TACCs is unlikely to be attainable or sustainable, nor would it allow the stock to move towards a size that will support the MSY” (Ministry of Fisheries 2008a). There is clearly a need to monitor the abundance of individual flatfish species.

1.1 Commercial fishery for flatfish

About half of the total landed flatfish weight in New Zealand is caught in FLA 3 (Ministry of Fisheries 2008a), an area that includes Fishery Management Areas (FMAs) 3, 4, 5, and 6 (Figure 1). Annual landings in FLA 3 fluctuate more than two-fold and have averaged about 1700 t since 1986–87 (Figure 2). The TACC has never been exceeded and has been between 44% and 96% caught, but overall has averaged 66% caught since 1999–2000.

The original TAC was set high because flatfish growth is fast and recruitment is variable, with the fishery relying on one or two year classes each year (Ministry of Fisheries 2008a). The high TACC has allowed fishers to take advantage of years when abundance was high (for example 1996–97), a management regime similar to that for red cod, a species which is also fast growing, has variable recruitment (Beentjes 2000) and similar cyclical peaks and troughs in catches. Although flatfish catches are not declining, there has not been a peak catch close to the TACC since 1996–97. It is difficult to gauge the effect of current fishing pressure on individual species and it is possible that although FLA 3 landings appear reasonably stable (accepting the inherent annual variability), individual species may be stressed. The eight species of commercial flatfish have distinct life histories with differences in distribution, growth rates, spawning behaviour, and biology (Colman 1994). To more effectively manage flatfish under the QMS, information on catches by individual species is required.

The bulk of the FLA 3 commercial catch (95%) is taken by inshore bottom trawl (CELR) with about 5% from setnetting (data from Beentjes 2003). Insignificant amounts of soles, brill, and turbot are taken as incidental bycatch by deepwater trawlers and reported on TCEPRs. The bottom trawl flatfish catch is taken from the inshore (less than 70 m) coastline of the east coast South Island and Southland. The bulk of setnetting catch is thought to be from Te Waihora (Lake Ellesmere), though it is not possible to determine the definitive amount because there is no reporting statistical area specifically for Te Waihora and all catch has been reported against area 022 (Banks Peninsula-Canterbury Bight). However, in Otago statistical areas 024 and 026, 99% of all landed flatfish was taken by bottom trawling, hence it is highly likely that the setnet flatfish catch in statistical area 022 is all from Te Waihora. Further, discussions with fishers suggest that setnetting has seldom been used to target flatfish in FLA 3. All eight commercial flatfish species are caught in FLA 3. The dominant species caught by trawling are lemon sole (LSO) and New Zealand sole (ESO), followed by sand flounder (SFL), with smaller landings of greenback flounder (GFL), brill (BRI), turbot (TUR), and yellowbelly flounder (YBF) (Beentjes 2003). In Te Waihora, flatfish setnet catches include black flounder (BFL) and YBF, with lesser quantities of SFL (Clem Smith, commercial fisher, pers. comm.).

4

1.2 Commercial fishing gear

Flatfish are caught in either standard multispecies trawls, or specialised flatfish target trawls. For target flatfish bottom trawling, specialised gear is required in which the heavy roller and bobbin ground rope of standard trawls is replaced by a chain fitted with short drop chains. A lighter chain known as the “tickler chain” in front of the chain ground rope disturbs flatfish from the bottom and herds them into the mouth of the net. The wings of flatfish trawls tend to have mesh sizes of less than 6“ (about 15 cm) compared to about 9” (about 23 cm) for standard trawl nets. Flatfish fishing also requires a low headline height of about 1 m (standard trawl is several metres), achieved by adjusting the flotation of the headline. Targeting flatfish using this gear can only be carried out on sandy bottoms because the gear is not designed to withstand the chaffing of rocks and rough bottom. Hence the gear is not used near river mouths and where sand flounder and yellowbelly flounder are more common. Flatfish are most commonly targeted between about 8 to 15 m depth, and not deeper than 50 m. Most vessels that target flatfish by trawl tend to make trips of only one to two days.

The standard trawl gear used in multispecies inshore fisheries of the east coast South Island and Southland is designed to target a range of demersal species that include red cod, barracouta, red gurnard, elephantfish, giant stargazer, tarakihi, spiny dogfish, sea perch, and dark ghost , whereas flatfish tend to be a relatively small bycatch. Flatfish are less vulnerable to the standard trawl gear than the target flatfish trawl gear, ostensibly because they tend to escape under the ground rope.

Target flatfish setnetting in Te Waihora normally uses 5.5” (14 cm) monofilament mesh. Nets are set in about 1500 m lengths (maximum length allowed) overnight.

1.3 Recreational and customary fishery

Flatfish and particularly the flounder species (SFL, YBF and BFL) that inhabit harbours, estuaries, coastal lakes, and shallow bays are important recreational species. While taken from throughout FLA 3 important areas include Te Waihora, Banks Peninsula, Otago Peninsula, Catlins, Oreti and Bluff estuaries, and lagoons on the Chatham Islands. Methods include setnetting, drag netting, and spearing (Ministry of Fisheries 2008a). The estimated recreational catch from a 1999–2000 survey (Boyd & Reilly 2005) was 395 000 (c.v. 33%) flatfish or 128 to 252 t. Assuming a tonnage midway between this range, then recreational fishing is about 10% of the average commercial catch. Although flatfish are also an important customary fishery, this catch is unknown.

1.4 Te Waihora (Lake Ellesmere)

Te Waihora is a shallow (maximum depth 2 m), brackish coastal lake just south of Banks Peninsula separated from the coast by a narrow 25 km long gravel bank known as Kaitorete Spit (Figure 3). The lake area is about 20 000 ha with 75 km of shoreline. The lake itself is about half the size of pre- European times when it was surrounded by extensive wetlands and forest. Freshwater input is from run-off from surrounding farmland, many small streams, springs, and the Selwyn and Halswell Rivers. The lake is often closed to the sea by wave action of coastal storms. When water levels rise above a specified level Environment Canterbury employ a mechanical digger to open a channel at Taumuta (southern end) to allow the lake water levels to fall — this has occurred from one to six times per year since 1945. This also allows inward flow of ocean water at high , presumably facilitating recruitment of fish species into the lake and migration out of the lake. The lake is defined as eutrophic and receives a high nutrient and sediment load from the surrounding farmland, which has been amplified by the loss of wetlands. The strong winds and shallow nature of the lake facilitate oxygenation of the water and prevent toxic algal blooms. The lake has two main commercial fisheries, tuna (freshwater ) and flatfish. The fishery is predominantly shortfin ( Anguilla australis ) with

5 smaller numbers of longfin ( A. dieffenbachii); fishers use fyke nets and the fishery season runs from about October to March. There is a small flatfish bycatch in eel fyke nets. The flatfish species in the lake includes black flounder, yellowbelly flounder, and sand flounder. The lake eel fishers switch to setnetting for flatfish in the winter months when eels are not vulnerable to capture. Black flounder have a limited distribution and tend to be confined to brackish estuaries. Te Waihora is ideally suited to black flounder and is the largest fishery in New Zealand for this species, while the other two flounder species are found coastally and within estuaries throughout New Zealand. Te Waihora is regarded as an important site for mahinga kai by local runanga, particularly tuna (eels). It is also an important recreational fishery for flatfish, trout, perch, and . The lake is well known for its diverse range of bird species and wildlife reserves.

1.5 Reporting flatfish catch

For management purposes, landings of all flatfish species are combined under the generic species code FLA and managed within the Quota Management System (QMS) essentially as a single species comprising four stocks (FLA 1, FLA 2, FLA 3, FLA 7). Hence, in FLA 3 all eight species of flatfish are landed in various quantities, but all are aggregated under the code FLA 3 for reporting against quota on QMRs/MHRs.

The fisheries reporting regulations require that only the code FLA be used in the Catch Landing section of the Catch Effort Landing Return (CELR). However, the codes for individual flatfish species (BRI, BFL, ESO, GFL, LSO, SFL, TUR, and YBF) are required to be used in the Catch Effort section of the form. In practice, however, MFish accept either individual species codes or the generic FLA code because of the difficulties it imposes on fishers to comply with the reporting requirements (Kim Duckworth, Ministry of Fisheries, pers. comm.). In 1989–90 when CELR and TCEPR replaced FSU forms, all estimated catches were reported as FLA, the percentage declining to about 40% by 1992–93 and remaining reasonably stable thereafter (Figure 4). The balance of the landings was recorded correctly using the individual species codes (Beentjes 2003) (Figure 4). In FLA 3 the proportion of catch reported as FLA was relatively stable at about 30% from 1993–94 to 2001–02.

Flatfish often do not make up the top five species by weight, and thus are not recorded in the catch and effort section of the CELR form. Estimated flatfish catch was about 85% that of landed flatfish catch from 1989–90 to 2001–02 (Beentjes 2003), indicating either under-estimation and/or the species was outside the top five species. A new reporting form (Trawl Catch Effort Return – TCER) introduced in 2007–08 for trawlers between 6 and 28 m in length, has provision to record the estimated catch and effort for eight species as well as latitude and longitude of tows. Hence we may see more flatfish species estimated catch and effort data recorded in the future.

As noted, Te Waihora does not have a specific statistical area and catches have been recorded against statistical area 022 on CELR forms. In 2006–07 a setnet specific form (Net Catch Effort Landing Return – NCELR) was introduced for vessels 6 m and over and has provision to record the estimated catch and effort for eight species as well as latitude and longitude of sets. Fishers with vessels under 6 m continue to use the existing CELR for both setnetting and trawling methods, which records only the top five species and statistical area, but not latitude and longitude.

Hence we now have flatfish catch effort potentially reported from two types of setnetting (CELR, NCELR) and three types of trawling forms (CELR, TCER, and TCEPR).

6 1.6 Biology of flatfish

1.6.1 Growth

The flounders and soles are fast growing short-lived species and the fisheries are thought to be comprised of only a few year classes. Sand flounder have been shown to grow to a maximum length of about 45 cm and reach a maximum age of about 6 years (Colman 1978) with the fishery comprising mostly two and three year old fish. Aging on New Zealand sole indicates a maximum age of 7 to 10 years at about 50 to 55 cm total length (Stevens et al. 2004). . Preliminary work on greenback flounder indicates that most fish in the commercial fishery are about 2 years old and the maximum age is about 6 years (NIWA unpublished data). Some ageing work on lemon sole aging has been carried out, but has not been published– there is reference to this in the plenary report (Ministry of Fisheries 2008b) where M is estimated at between 0.62 and 0.96. There are no published studies on the age and growth of yellowbelly flounder and black flounder.

The two colistium species, brill and turbot, however, grow much larger and older, growing rapidly in the first few years after which growth is slow (Stevens et al. 2005). Stevens et al. (2005) found brill maximum age to be 21 years with the largest fish about 47 cm, whereas brill are known to grow as large as 70 cm indicating a possible maximum age older than 21 years. Similarly the oldest turbot was 16 years and the largest fish 62.5 cm but turbot are known to grow as large as 80 cm.

1.6.2 Distribution, movements, and spawning

Flounder species are generally found in shallower waters and can often be caught near river mouths, whereas soles, turbot and brill are found coastally and as deep as about 70 m. The harbours and estuaries of Otago Peninsula are nursery grounds for juvenile sand flounder and greenback flounder (Roper & Jillett 1981). Similarly, Banks Peninsula harbours and estuaries are nursery grounds for sand flounder (Colman 1978). Colman (1974) showed that sand flounder and yellowbelly flounder in the Firth of Thames exhibit very limited movement with a winter-spring spawning migration restricted to the northern part of the bay around Waiheke Island. Conversely, Colman (1978) found that sand flounders tagged within the harbours of Banks Peninsula tended to disperse from the nursery grounds out into Pegasus Bay and southward into the Canterbury Bight, moving further from the tagging sites with time and as they grew larger. Spawning is also thought to take place in winter-spring with two main spawning grounds off Akaroa and Timaru in about 30 to 45 m depth — the eggs and larvae are then carried by the northward flowing currents settling in, among other areas, the harbours of Banks Peninsula. Hence, while the Firth of Thames sand flounder may have very restricted stock boundaries, on the east coast of the South Island, the stocks may be less localised. It seems likely that an area as large as the east coast South Island and Southland will have multiple localised stocks with minimal mixing.

1.7 Monitoring flatfish abundance

The only fishery-independent abundance indices for flatfish in FLA 3 are from the ECSI Kaharoa trawl surveys that have included North Otago, Canterbury Bight, and Pegasus Bay areas (30–400, winter surveys; 10–400 m summer surveys) (see Stevenson & Beentjes 2001, Beentjes & Stevenson 2008). However, the Kaharoa trawl gear is not designed to target flatfish, and the winter surveys are generally too deep for some species. Hence, flatfish species are not considered to be adequately monitored by the Kaharoa trawl surveys. They may be useful, however, for comparative purposes to examine general changes in relative biomass and distribution among the surveys.

Catch effort analyses may provide relative abundance estimates for flatfish, but until recently, had not been attempted, probably because of the number of species involved and the uncertainty surrounding reporting of estimated catches. The suitability of using catch effort data to monitor flatfish stocks was

7 addressed by Beentjes (2003) who described the nature and extent of available flatfish data, and made recommendations on whether it was feasible to carry out catch effort analyses. Beentjes (2003) concluded that there was merit in conducting CPUE analyses for key flatfish species from each QMA, which had been reported by species in the catch effort section of CELR forms. Further, analyses on the species code FLA would be questionable as it comprises multiple species and trends for all species are unlikely to be similar. A standardised CPUE analysis was subsequently attempted for YBF and SFL, the two key flatfish species in the FLA 1 setnet fishery (Coburn & Beentjes 2005). Analyses were carried out for both species combined and individually in FLA 1, showing the value in carrying out CPUE analyses for individual flatfish species, where data are adequate.

The fishing industry collected some flatfish species catch-effort and biological data in the southeast trawl fishery during elephantfish sampling, a monitoring requirement under the adaptive management programme (AMP). Data were collected voluntarily and recorded on the SeaFIC logbook reporting form in 2004–05, 2005–06, and 2006–07.

1.8 Objective

To characterise the FLA 3 fishery and analyse CPUE trends in the fishery using CELR data up to the end of 2006-07.

2. METHODS

2.1 Catch effort data extract

Data were extracted from the Ministry of Fisheries catch effort database from CELRs and TCEPRs for 1989–90 to 2006–07, and NCELRs for 2006–07. CELR forms are completed by smaller (under 28 m length) inshore domestic vessels and record the daily estimated catch and associated effort for the top five species (by weight) for individual tows/sets (catch effort section) and the actual landed weight (catch landing section), when the catch is landed (i.e., for each trip). TCEPRs are used by trawlers over 28 m and also record the estimated catch of the top five species, but record tow by tow estimated catch, and landed catch is recorded on separate CLRs. The NCELRs are used by setnetting vessels under 6 m length. The new TCER form for inshore trawlers was not introduced until 2007–08 and hence no extracts were made from this form.

The extract included data from all ‘fishing events’ and ‘landing events’ from trips where landings of FLA 3 (i.e., fishing event from statistical areas 18–32, 49–51, 301–303, 401–412, 501–503, and 601– 625) were reported between 1 October 1989 and 30 September 2007. The fish stock (FLA 3) and landed weight were taken from the catch landing section of the CELR and NCELR and from the CLRs for TCEPRs and linked to the effort data by the MFish trip_number variable. Our selection included all bonafide flatfish species codes (BFL, BRI, ESO, GFL, LSO, SFL, TUR, and YBF) as well as the generic FLA code and incorrect but sometimes used codes SOL, BLF, and FLO.

2.2 Catch effort data grooming and restratification (roll-up)

The catch effort data were groomed and restratifed using Starr’s (2007) data processing method as implemented by Manning et al. (2004) with refinements by Blackwell et al. (2005), Manning (2007), and from NIWA unpublished results from MFish research projects SPD2005-01 and SCH2006-01. The set of relevant or “candidate” fishing trips (i.e., non-null, non-zero, positive landing of FLA 3 reported between 1 October 1989 and 30 September 2007 and all fishing and landing events associated with these trips) was treated as follows:

8 The basic unit of data within the algorithm is the fishing trip. The major steps were as follows.

Step 1: The fishing effort and landings data were first groomed separately. Outlier values in each variable that failed a range check were corrected using median imputation. This involved replacing missing or outlier values with a median value calculated over some subset of the data (e.g., the number of trawls per day). While this may lead to underestimation of the variance for a given variable, this uses the data to “fix itself” rather than merely dropping cases containing missing or outlier data, maximising the amount of data available for analysis while eliminating missing or implausible values.

Step 2: The fishing effort within each valid trip was then restratified (i.e., statistical area, method, trip, and target species).

Step 3: The greenweight landings for each fishstock for each trip were then allocated to these effort strata using the relationship between the statistical area for each effort stratum and the statistical areas contained within each fishstock.

Step 4: The greenweight landings were then allocated to the effort strata using the total estimated catch in each effort stratum as a proportion of the total estimated catch for the trip. If estimated catches were not recorded for the trip although a landing was recorded for the trip, then the total fishing effort in each effort stratum as a proportion of the total fishing effort for the trip was used to allocate the greenweight landings.

Processed product weights in New Zealand fisheries are converted to greenweight catches using species and product-form-specific conversion factors (multiplicative constants). Flatfish are always landed green, hence conversion to green weight was not required.

2.3 Catch effort data descriptive analyses

Characterisation analysis was carried out using the groomed, restratified, and merged datasets. We first explored the flatfish data and related the aggregated flatfish coded catch (i.e., all aggregated flatfish codes) to associated effort strata month, location (statistical area), method, and target species using a bubble plot. We then explored the data in more detail by plotting flatfish catch (aggregated flatfish codes) against combinations of the variables year, month, statistical area, method, target species, and form type.

Landed catch was compared with estimated generic flatfish catch (FLA) because fish stock (FLA 3) is recorded as FLA and not by individual species. This provided an estimate of the accuracy of estimating catch and also the extent to which flatfish are not recorded in the top five species.

Vessel composition was examined to determine the extent to which vessels maintained a presence in the fishery and any changes in vessel size over time. This was used to assist in the selection of ‘core vessels’ for inclusion in the standardised CPUE analysis.

The descriptive data were inspected for trends that might represent changes in fishing practices to assist us with interpreting the standardised CPUE indices.

2.4 Estimating species-specific (BFL, BRI, ESO, GFL, LSO, SFL, TUR, and YBF) catch histories from the aggregated FLA 3 data

There are eight individual species administered within the single generic code FLA, but we are interested in calculating species-specific catch histories, and species-specific relative abundance indices from the corresponding standardised CPUE.

9

To deal with the problem of having much of the estimated catch recorded under the generic code FLA, we estimated the proportion of each flatfish species in the catch and then applied this to the catch recorded by the generic FLA codes to yield species-specific catch estimates. We assumed necessarily that the FLA landed catch is an accurate index of the true level of catch of each of the eight species of flatfish combined in FLA 3.

The individual unknown species-specific proportions were estimated from two sources. 1) The associated non-generic estimated catches for the associated fishing records (algorithm score “2”), or trip (algorithm score “4”). 2) Sanford Ltd Timaru dataset of species-specific landing weight data collected over the periods 1992–93 to 2000–01 and 2003–04 to 2007–08. These are catches from trawling only.

The algorithm applied to the aggregated flatfish merged and groomed dataset is shown in Figure 5.

2.5 Standardised CPUE analyses

Because the setnet fishery is almost exclusively confined to Te Waihora and the trawl fishery to the inshore coastal waters (excluding Te Waihora), trawl and setnet flatfish fisheries are dealt with separately.

Standardised CPUE indices were obtained by fitting Generalised Linear Models (GLMs) (McCullagh & Nelder 1989) to the groomed, restratified (merged), and imputed (generic FLA codes converted to species-specific codes) dataset derived from the characterisation dataset using the methods of Dunn (2002) and Manning et al. (2004). In each case, the response variable was non-zero log(catch) and the relevant effort variables (setnet or trawl) were used as predictors in the models. Two sets of CPUE analyses were carried out, one for the bottom trawl fishery and another for the Te Waihora flatfish setnet fishery.

A forward, stepwise, multiple-regression fitting algorithm (Chambers & Hastie 1991) implemented in the “R” statistical programming language was used to fit all models. The algorithm generates a final regression model iteratively and uses a simple model with a single predictor variable, fishing year , as the initial or base model. The reduction in residual deviance relative to the null deviance, R2 , is calculated for each additional term added to the base model. The term that results in the greatest reduction in residual deviance is added to the base model if this results in an improvement in residual deviance of more than 1%. The algorithm repeats this process, updating the model, until no new terms can be added. A stopping rule of 1% change in residual deviance was used as this results in a relatively parsimonious model with moderate explanatory power.

Following Manning et al. (2004), the following predictor variables were offered to the trawl models: fishing year , statistical area , month , target species , vessel key , fishing duration , effort number, effort width, effort height . For the setnet models the variables were fishing year , month , target species , vessel key , fishing duration , effort setnet (net length), effort width, effort width (mesh size). All continuous predictor variables were offered to the models as third-order polynomial functions. The number of effort variables that may sensibly be offered to the models is limited by the resolution of the effort strata (i.e., unique vessel-statistical area-method-target species combinations). The distributions of all predictor variables were investigated before the model fits to assess whether changes in these distributions had occurred.

Each model was fitted to a “core vessel” subset (vessels with a “consistent” presence in each dataset). Core vessels were defined as vessels with a presence of five years or more in the fishery for the trawl model, and three years for the setnet model, with both having at least 10 landings per year.

10 Following Dunn (2002), all model indices are presented in a canonical form. Model fits are investigated using standard regression diagnostic plots. For each model, a plot of residuals against fitted values and a plot of residuals against quantiles of the standard normal distribution are produced to check for departures from the regression assumptions of homoscedasticity and normality of errors in link-space. Plots of the expected catch rate for each variable in the final model holding all other variables fixed at median values are also produced. We investigated whether the CPUE indices are monitoring abundance using the criteria developed by Dunn et al. (2000). This involved examining the relationship between CPUE and abundance, considering data adequacy, model fit, and model validation. For the latter, we reviewed species-specific flatfish relative biomass estimates and survey size- composition data from research trawl surveys of the east coast of the South Island by Kaharoa during the winter and summer survey time series (see Section 2.5.1). We also reviewed the New Zealand Seafood Industry Council (NZ SeaFIC) logbook size composition data for flatfish (see Section 2.5.1).

2.5.1 Ancillary data

R.V. Kaharoa surveys Estimated biomass, distribution of catch rates, and length frequencies of the flatfish species from the R.V. Kaharoa east coast South Island summer (1996 to 2000, 5 surveys) and winter (1991–1994, 1996, 2007–2008, 7 surveys) time series inshore trawl surveys were extracted from the trawl database. Only LSO, ESO, and SFL are caught on these surveys in any number, and then only in relatively small quantities. Biomass estimates tend to have very large coefficients of variation (c.v.s) reflecting the low catches and few tows that they are present. To show the relative proportions of flatfish species caught on these surveys the mean proportion of biomass for each species across winter and summer surveys respectively was calculated, and the 95% confidence intervals calculated to indicate how variable biomass was among the surveys.

Processor’s data Data on individual flatfish species trawl fishery landings in FLA 3 processed by Sanford Ltd Timaru for 1992–93 to 2007–08 (with missing data from 2001–02 and 2002–03) were provided to NIWA upon request. These data were plotted by flatfish species catch by year and percent species catch by year for the main species.

Industry logbook programme SeaFIC provided data on flatfish catch sourced from a logbook scheme run in 2004–05, 2005–06, and 2006–07 in FLA 3, ostensibly to monitor elephantfish (ELE 3) under the AMP. Data were collected throughout FMA 3 voluntarily by 10 vessels based in Bluff, Oamaru, Timaru, and Kaikoura. As well as estimating the catch weights of individual flatfish species, length frequency measurements were taken from BRI, ESO, LSO, SFL, TUR, and YBF. These data were collected sporadically, when time allowed.

3. RESULTS

3.1 Catch effort data descriptive analyses

3.1.1 Merged data diagnostics

The relationship among key quantities in the groomed and merged datasets are shown in Figure 6. There is a discrepancy between the QMR catch and the landed greenweight catch from our merged dataset in the mid 1990s. Overall the estimated FLA catch seems to be typically less than half of the corresponding landed greenweight catch recorded on landing forms and presumably counted against quota. There are a large number of zeros in the data, indicating that many vessels have returned a landed greenweight record but no corresponding estimated catch in their corresponding effort records (Figure 7).

11 Of the merged data, the destination for 98% of landed catch was reported as ‘Landed in New Zealand to a licensed fish receiver’ (Table 1). Removal of the destination codes that could have potentially resulted in double-counting accounted for only 1.5% of the total dataset, by weight.

The conversion factor type, applied to the various landed states using the most recent valid value specified for each product state code in the groomed and merged data is shown in Table 2.

3.1.2 Fishery characterisation

The merged and groomed aggregated catch data are expressed in various combinations of the variables year, month, form type, method, target species, and statistical area caught (see below).

General summary For all aggregated flatfish species codes, catches by year and month, year and statistical area, year and method, and year and target species are summarised in a single bubble plot (Figure 8). A further breakdown by year, month, and method is shown in Figure 9. Catches show no strong seasonal trends, but overall tend to be highest from October to March. The important statistical areas are 020, 022, 024, and 026, all of which are adjacent to the east coast South Island coastline. Flatfish are taken predominantly by bottom trawl with lesser amounts by setnet (Figures 8 and 9). Most of the flatfish catch has been by vessels for which the generic code FLA was recorded as the target species with modest catches of red cod as the target; there is some catch taken for five individual flatfish species recorded as the target, but most of this is from ESO and LSO target.

Detailed breakdown A detailed breakdown of aggregated flatfish catches by method, year, and form type indicates that in the flatfish trawl fishery, virtually all landings were from vessels completing CELR forms with negligible landings from TCEPRs (Figure 10). There are virtually no landings from the NCEL form because this form came into effect only in 2006–07 and only for vessels over 6 m. A further breakdown of form type by target species and method, shows clearly that the key target species for trawling is the generic code FLA, and to a lesser extent RCO (Figure 11). A breakdown of form type by statistical area and method indicates that the setnet method was used exclusively in statistical area 022, the reporting area used by Te Waihora fishers (Figure 12); this is strong evidence that the catches of flatfish by setnet is almost exclusive to Te Waihora.

A detailed breakdown of aggregated flatfish catches by target species and statistical area show that the trawl flatfish catch with FLA as the target is spread throughout statistical areas 020, 022, 024, 025, 026, and 030, and with red cod as the target, in areas 020 (Pegasus Bay) and 022 (Canterbury Bight), the main red cod fishing grounds (Figure 13). When targeting the individual flatfish species LSO, ESO, and SFL by trawl, the catches are only low to modest across coastal statistical areas 018 to 030, but are confined to 022 for setnet target SFL, BFL and YBF; further evidence that the flatfish setnet fishery is confined to Te Waihora and that the target species are SFL, BFL, and YBF.

Estimated species-specific catch histories from the aggregated FLA 3 data Inputted catch histories for individual flatfish species after applying the species catch estimation algorithm (see Figure 5) to our merged and groomed dataset are shown in Figure 14. The resulting catch histories do not change the conclusions of the fishery characterisation carried out on the original merged and groomed dataset that included the aggregated flatfish codes. The only three flatfish species caught in any quantity by trawling are ESO, LSO, and to a lesser extent SFL. Catches of these species are spread throughout the coastal statistical areas from 020 to 030, although LSO catch was greatest in 026 (Figure 14).

The setnet catch again is confined to statistical area 022 (= Te Waihora) and the only species caught in any amount are BFL, SFL, and YBF.

12

3.2 CPUE analyses

3.2.1 Bottom trawl fishery

CPUE model datasets used in the bottom trawl fishery analyses are shown in Table 3. Datasets were defined by method, target, statistical area, response variable, catch score from the algorithm, and core vessel restrictions. In all models, for core vessels method was bottom trawl, target was the mixed species (BAR, ELE, ESO, FLA, GUR, LSO, RCO, SFL, SPD, SPE, STA, TAR, & YBF ), statistical areas were 020, 022–024, 026, 030, and response variables were log(catch) of species FLA, ESO, LSO, and SFL. Two models (all scores and score “4” only) were run for each species, resulting in eight models (0.1 to 0.8).

The proportion of zeros in each of the eight datasets is shown in Figure 15 and indicates the level of presence/absence of each species in trips. The unstandardised and standardised CPUE indices with 95% confidence intervals are shown in Figures 16–19. The expected catch rates for predictor variables included in the models, model diagnostic plots, and vessel history plots are shown in Appendix 1.

For all trawl CPUE models (0.1 to 0.8) there is considerable variation in the CPUE among years with indications of large scale swings in abundance within a few years. The confidence intervals surrounding the annual CPUE indices are generally low, indicating that there is strong evidence to support the observed trends.

For the FLA models (Figure 16) which include the aggregated flatfish species, the standardised and non- standardised indices have similar overall trends. For both models (‘all scores’ and ‘score 4 only’) standardised CPUE increased from the early 1990s to the mid 1990s, and then declined until 2005–06, with indications of an upturn in 2006–07.

For the ESO models (Figure 17), the standardised and non-standardised indices have similar overall trends, particularly the ‘score 4 only’ model; for the ‘all scores’ model, standardised CPUE increased from the early 1990s to 1996–97, declined steeply over the next two years, increased slightly through to 2002–03, and declined steeply again to 2006–07. The ‘score 4 only’ model has much the same trend except that the second CPUE peak is greater than the first.

For the LSO models (Figure 18), the standardised and non-standardised indices have similar overall trends; for the ‘all scores’ model standardised CPUE increased from the early 1990s to 1997–98, and then declined steeply to 2003–04, before a steep increase to 2006–07. The ‘score 4 only’ model has much the same trend although it tends to be flatter with a more accentuated peak in CPUE between 1997–98 and 1998–99.

For the SFL models (Figure 18), the standardised and non-standardised indices have similar overall trends; for the ‘all scores’ model standardised CPUE increased from the early 1990s to 1995–96, and then declined steeply to 1999–2000, before generally increasing to 2006–07, with the exception of 2005–06 which dropped suddenly. The ‘score 4 only’ model has much the same trend although it tends to be flatter and the CPUE peak is a year earlier in 1994–95.

3.2.2 Setnet fishery

CPUE model datasets used in the Te Waihora setnet fishery analyses are shown in Table 4. Datasets were defined by method, target, statistical area, response variable, and catch score from the algorithm and core vessel restrictions. In all models for core vessels, method was setnet, target was the mixed flatfish species (BFL, FLA, SFL YBF), statistical area was 022, and response variables were log(catch) of species FLA, ESO, LSO, and SFL. Two models (‘all scores’ and ‘score 4 only’) were run for each species, resulting in eight models (0.9 to 0.16). Because the Sanford’s data did not have any landings of YBF and BFL, by

13 definition scores 1 and 2 were seldom used to estimate species-specific catch and hence there is very little difference between the data used in the ‘all scores’ and ‘score 4 only’ models (see Figure 5).

The proportion of zeros in each of the eight datasets is shown in Figure 20 and indicates the level of presence/absence of each species in trips. The unstandardised and standardised CPUE indices with 95% confidence intervals are shown in Figures 21–24. The expected catch rates for predictor variables included in the models, model diagnostic plots, and vessel history plots are shown in Appendix 1.

For all setnet CPUE models (0.1 to 0.8) there is considerable variation in CPUE among years with indications of large scale swings in abundance between years. The confidence intervals surrounding the annual CPUE indices are greater than for the trawl analyses, indicating that evidence to support the observed trends is sometimes weak.

For the FLA models, which include the aggregated flatfish species, the standardised and non-standardised indices have similar overall trends (Figure 21). For both models (‘all scores’ and ‘score 4 only’) standardised CPUE fluctuates greatly with four distinct peaks in 1990–91, 1995–96, 2001–02, and 2004– 05. No clear or consistent trend in CPUE is apparent, although CPUE is greater, on average, after 1995.

For the BFL models the standardised and non-standardised indices have similar overall trends (Figure 22). For both models (‘all scores’ and ‘score 4 only’) standardised CPUE fluctuates greatly with four distinct peaks in 1990–91, 1995–96, 2001–02, and 2004–05. No clear or consistent trend in CPUE is apparent, although CPUE is greater, on average, after 1995.

For the SFL models the standardised and non-standardised indices have similar overall trends (Figure 23). For both models (‘all scores and ‘score 4 only) standardised CPUE indices fluctuate greatly between years. The CPUE indices from the ‘all scores’ model (model 0.13) that includes Sanford’s data to estimate catch, differs from the ‘score 4 only’ model. The ‘all scores’ model has two large peaks in 1990–91 and 1996–97, but has shown a steady increase in CPUE since 1999–2000. The ‘score 4 only’ model shows no clear trend over time but does have strong peaks in 1990–91 and 1996–97 similar to the ‘all scores’ model.

For the YBF models the standardised and non-standardised indices have similar overall trends with the exception of 1995–96 in which the standardised model has a strong peak (Figure 24). For both models (‘all scores’ and ‘score 4 only’) standardised CPUE indices are virtually identical and fluctuate greatly between years with no clear trends.

3.3 Ancillary data

3.3.1 Sanford flatfish bottom trawl landings data

The most common commercial flatfish species in FLA 3 landed into Sanford are lemon sole (LSO, 39%) and New Zealand sole (ESO, 32%), followed by sand flounder (SFL, 15%) (Table 5, Figure 25). The Sanford data, on average, account for about 12% of FMA 3 commercial landings and will be biased by the size of vessels landing the catch and whether the flatfish is bycatch or a target species; we have no information on vessel size or target species associated with the landed flatfish, but given that most vessels landing into Sanford are trawlers, and the results of the characterisation, it is likely that these fish are virtually all trawl caught. While some processors have recorded as much as 12% of annual flatfish landings as black flounder (BLF) (2003), this is almost exclusively from Te Waihora (Lake Ellesmere), which is why it is not caught on trawl surveys or in the catch processed by Sanford.

The Sanford data indicate that ESO landings increased markedly between 1994–95 and 1998–98 with a peak catch in 1996–97, with another spike in 2004–05 (Figure 25). Similarly LSO landings peaked the following year in 1997–98 and, notwithstanding the missing years, showed a steep increase from 2003–04 to 2007–08. Landings of other species fluctuated between years with no trends.

14

3.3.2 SeaFIC logbook programme

About 39 t of flatfish were included in the logbook programme data. New Zealand sole (LSO) was by far the main species caught (61%), followed by LSO (11%) and SFL (23%) (Table 5). These catches were mainly by vessels targeting elephantfish inshore in shallow water, probably in less than 30 m. The high proportion of ESO and SFL in the catch is consistent with the Kaharoa trawl survey shallow water tows (10–30 m) (Table 5) in which ESO and SFL are more common than offshore, where LSO is the main species.

The length frequency distributions are shown for 557 BRI, 3853 ESO, 2304 LSO, 2753 SFL, 12 TUR, and 124 YBF (Figure 26). Of the two sole species, ESO is considerably larger than lemon sole and has a wide size distribution with a small mode at around 61 cm total length. Sand flounder also has some very large fish in the catch, with some recorded up to 63 cm total length. Distributions of LSO, ESO, and SFL suggest the presence of more than one mode, indicative of cohort modal progression.

3.3.3 Kaharoa ECSI trawl survey data

The Kaharoa trawl gear used is not designed to target flatfish (Parker et al. 2009) and it is unlikely to go shallow enough to properly survey sand flounder, particularly in the winter surveys when the minimum depth range is 30 m. The trawl gear has a 5 cm gap between the groundrope and the fishing line that could result is escapement of flatfish. Fishing for flatfish requires specialised gear with a chain groundrope and tickler chains (see Introduction). Hence, although we present biomass, catch rates, and length frequency distributions for flatfish species caught on the ECSI Kaharoa trawl surveys. the results need to be interpreted with these qualifications. It is possible the relative proportions of species are reasonably well represented within the depth ranges surveyed.

Biomass The mean proportions of flatfish biomass from the ECSI trawl surveys in 30 to 400 m indicate that lemon sole is the main flatfish species followed by ESO and SFL in winter, and ESO and GFL in summer (Figure 27, Table 5). Although TUR, BRI, YBF, and GFL were caught in 30–400 m, their numbers were insignificant. In shallow water during winter (10–30 m, 2007 winter survey), ESO and SFL were the key species, and LSO was no more common than BRI, TUR, and YBF. The species proportions were similar for the inshore (10–30 m) summer surveys, although greenback flounder was more common than in winter. Hence, there is a clear difference in the species mix inshore and offshore with LSO dominating catches offshore (30–400 m), and ESO and SFL dominating catches inshore (10– 30 m). Seasonal differences are less marked.

Biomass indices of the main species ESO, LSO, and SFL for the winter and summer time series surveys are shown in Figure 28. The winter surveys are not continuous and have some gaps (2 years between 1994 and 1996 and 11 years between 1996 and 2007). Winter biomass of each of these three species in 30–400 m fluctuate and show no consistent trend, although the 1991 ESO biomass was nearly four-fold greater than in following years. LSO and SFL biomasses during summer in 10–30 m fluctuate and show no trends, but indicate that for ESO there was a decline in 1999 and 2000 (Figure 28). ESO and LSO show clear declines in biomass from 1996 to 2000 in the summer surveys in 30–400 m, and SFL to a lesser extent.

15 Length frequency distributions There were only sufficient data to plot meaningful length frequency distributions for LSO, ESO, and SFL in the Kaharoa winter trawl surveys and these data were combined for all surveys (Figure 29). The LSO distribution has a clear mode at 25 cm and sizes range from 10 to 45 cm. New Zealand sole (ESO) has a similar size range, but no clear mode and tends to have more larger fish.. Sand flounder (SFL) has no clear modes and ranges in length from about 15 to 50 cm. In all three species there were more females caught than males, and the females grow larger than males.

Similarly for the summer Kaharoa trawl surveys, there were sufficient data to plot meaningful length frequency distributions only for LSO, ESO, SFL, GFL, and YBF (Figure 30). The summer survey length frequency distributions of LSO, ESO, and SFL are remarkably similar to those in the winter surveys. Greenback flounder (GFL) ranges in length from about 17 to 54 cm with indications of a juvenile and an adult mode. Yellowbelly flounder (YBF) distribution has a narrow size range with most fish between 30 and 50 cm. Most flatfish in the summer surveys were unsexed, so it is not possible to comment on the difference in size by sex.

Catch rates Catch rate distributions are shown for ESO, LSO, and SFL for the ECSI winter and summer Kaharoa trawl survey time series (Figures 31–36).

New Zealand sole (ESO) was caught in between 4 and 16% of tows (average 8%), in the winter surveys (depth range 30–400 m), generally in the shallowest tows throughout the survey area with no temporal changes in distribution over the time series (Figure 31). In 2007, when the 10–30 m depth was surveyed, ESO was caught in 54% of tows in this depth range. In the Kaharoa summer trawl survey time series (10–400 m), ESO was caught in between 9 and 41% of tows (average 24%), predominantly in the shallowest strata (10–30 m) throughout the survey area (Figure 32). There were no temporal changes in distribution over the time series, although catch rates were extremely low in 1999 and 2000 compared with the previous surveys in 1996 to 1998.

Lemon sole (LSO) was caught in between 27 and 59% of tows (average 48%) in the winter surveys (depth range 30–400 m) across the shelf, but catch rates were much greater in shallower tows throughout the survey area with no temporal changes in distribution over the time series (Figure 33). In the Kaharoa summer trawl survey time series (10–400 m) LSO was caught in between 21 and 65% of tows (average 45%), throughout the survey area in the same general depth range as the winter surveys (Figure 34). There were no temporal changes in distribution over the time series, although like ESO, catch rates of LSO were extremely low in 1999 and 2000, compared with the previous surveys in 1996 to 1998.

Sand flounder (SFL) was caught in between 1 and 9% of tows (average 6%) in the winter surveys (depth range 30–400 m) in the shallowest tows throughout the survey area with no temporal changes in distribution over the time series, although catch rates of SFL were much greater in 2007 and 2008, compared with the previous surveys from 1991 to 1996 (Figure 35). In the Kaharoa summer trawl survey time series (10–400 m) SFL was caught in between 7 and 26% of tows (average 16%), predominantly in the shallowest strata (10–30 m) throughout the survey area (Figure 36). There were no temporal changes in distribution over the time series although, like ESO and LSO, catch rates of SFL were extremely low in 1999 and 2000, compared with the previous surveys in 1996 to 1998.

16

4. DISCUSSION

The objective of this project was to characterise the FLA 3 fishery and analyse CPUE trends in the fishery. In addition, we also investigated ancillary data that included processor’s landings data, the SeaFIC logbook programme, and ECSI Kaharoa trawl surveys.

4.1 Characterisation analyses

Because flatfish landings comprise up to eight species, and much of the estimated catch and all the landed catch are reported under the generic code FLA, characterisation of this fishery is problematic. The key features of the aggregated flatfish species fishery in FLA 3 include: 1) no strong seasonal trend in catch; 2) important statistical areas are the those contiguous with the coast (020, 022, 024, and 026); 3) virtually all catch outside Te Waihora is taken by bottom trawl, and setnet within Te Waihora; 4) most of the flatfish catch was by vessels recording FLA as the target species with modest catches from target red cod, and lesser catches with target ESO and LSO; 5) virtually all flatfish catch is landed from vessels completing CELRs (up to 2006–07); 6) only three flatfish species are caught in any quantity by trawling (ESO, LSO, and to a lesser extent SFL); catches of these species are spread throughout the coastal statistical areas from 020 to 030, although LSO catch was greater in 026; 7) the setnet catch is virtually all from statistical area 022 (assumed to be Te Waihora), and the only species caught in any amount are BFL, SFL, and YBF.

The Kaharoa ECSI trawl survey catch rate data, although collected from surveys that used trawl gear that is not designed to target flatfish, still provide information on the relative biomass, as well as spatial and temporal distributions of the three key trawl species ESO, LSO, and SFL. Lemon sole is the main flatfish species, followed by ESO and SFL in winter, and ESO and GFL in summer (see Table 5). The catch rates indicate that there are distinct depth ranges for these species with SFL inhabiting the shallow coastal areas down to, and without doubt, inside 10 m. ESO are found deeper, predominantly in 10 to 30 m, with LSO extending deeper again into the mid, and occasionally, outer continental shelf. The length frequency distributions from the Kaharoa and industry logbook scheme are similar and there are some indications from the latter of modes that represent age classes (see Figure 26). There are insufficient data to plot length frequencies by survey to look at recruitment strength.

The Sanford data, which include landings from throughout FLA 3 and in a wide range of depths, also indicate that the most common flatfish species landed, in order, are LSO, ESO, and SFL (see Table 5). The industry elephantfish logbook scheme tended to collect flatfish data from shallow water (less than 30 m) where elephantfish are targeted, and consequently the flatfish species order differed from the Sanford data with ESO the most common followed by SFL and LSO. This is consistent with the depth distributions of these three species.

4.2 CPUE analyses

The CPUE analyses were initially carried out for the aggregated flatfish species, followed by the individual species ESO, LSO, and SFL for bottom trawl, and BFL, SFL, and YBF for the setnet fishery, with the species catch imputed from the algorithm. Any trend in the aggregated species is an amalgam of the trend of the three individual species.

For CPUE to be a useful tool for managing a fishery there is the necessary assumption that CPUE is proportional to abundance of the stock. Hence, we investigated whether the CPUE indices are monitoring abundance using the criteria developed by Dunn et al. (2000). This involved examining the relationship between CPUE and other proxies of abundance.

17 Trawl fishery The trawl fishery species models sometimes varied between those that used ‘all scores’, which included the Sanford data to prorate non-specific catch, and those that used ‘score 4 only’, in which catch was estimated from the species catch recorded from each trip (see Figure 17). This is not surprising since the Sanford’s catch have missing data for the years 2001–02 and 2002–03, resulting in an average value being used for these years. An implicit assumption is that the Sanford catch is proportional, in some way, to the abundance of each species. In reality, there may have been other factors in play such as market forces, fleet characteristics, and locations fished, target species, and possibly high grading which may occur as quota is allocated for the amalgamated species mix. However, we have no evidence to suggest that either model is superior and both are considered to be equally plausible. Some indices increase or decrease substantially from one year to the next, probably because flatfish have only a few year classes in the fishery and abundance is highly recruitment driven with large fluctuations in abundance between years. These large scale fluctuations in annual catch are also characteristic of the red cod fishery on the ECSI — red cod is also short-lived, and highly recruitment driven (Beentjes & Renwick 2001).

There is a clear correlation between the reported generic flatfish landed catch in FLA 3 (Ministry of Fisheries 2008a) and the trawl fishery CPUE indices for amalgamated flatfish (models 0.1 and 0.2), generated in this report (Figure 37). Both CPUE and landed catch increase and peak about 1997–98 before a steady decline, with indications of an upturn in 2006–07. Similarly, the summer ECSI Kaharoa trawl survey biomass of combined flatfish species indicates a decline in biomass after 1998, whereas the winter time series is missing too many years to make any comment on the relationship between CPUE and biomass (Figure 38). The Sanford flatfish landed data tend to mirror the trends in the FLA 3 landed catch and aggregated flatfish CPUE models (Figure 39). Based on this, there is evidence that CPUE is tracking abundance of the combined suite of flatfish species in FLA 3, and hence in some way is proportional to abundance, but this is driven predominantly by the main species, LSO and ESO.

The three key trawl species, ESO, LSO, and SFL show generally similar trends of increasing CPUE in the mid to late 1990s, followed by a steep decrease (see Figures 17–19). After this the three species differ and ESO has a secondary peak about 2001–02 followed by a steep decline, LSO increases again after about 2003–04, and SFL fluctuates with no clear trend. ESO CPUE displays a similar trend to the Sanford ESO catch data, and to a lesser extent to the ECSI summer Kaharoa trawl survey ESO biomass which shows declines in relative biomass in 1999 and 2000 (see Figures 17, 25, and 28). Similarly, LSO also displays a clear relationship among CPUE, Sanford LSO catch data, and the ECSI summer Kaharoa trawl survey LSO biomass, but in the 30–400 m depth range where it is most common (see Figures 18, 25, and 28). SFL displays a clear relationship between CPUE and Sanford’s LSO catch data and to lesser extent to the summer trawl survey biomass (see Figures 19, 25, and 28). Thus for all three species there is evidence that the CPUE indices generated for these three species are a proxy for abundance.

Te Waihora setnet fishery The Te Waihora setnet CPUE models (‘all scores’ and ‘score 4 only’), are virtually identical for BFL and YBF because Sanford data do not include these species and hence the imputed catch was similar for the two models. Further, the SFL model for ‘all scores’ (model 0.13) is therefore not strictly valid because the Sanford SFL catch was taken outside Te Waihora and the model 0.14 (score 4 only) should be used. The setnet CPUE models are marked by large fluctuations in the annual CPUE indices with no clear trends, and wide confidence limits around the indices, the latter a reflection of having only about seven participants in the fishery. Unlike the trawl CPUE indices, there are no proxies of abundance to compare with the setnet CPUE indices to test for plausibility.

Since 1987 the lake has been opened either naturally or artificially every year from one to six times and the period open has varied from 3 to 97 days. Black flounder is confined to brackish waters and we do not know if there is any spawning-related migration in or out of the lake when the lake is open. The frequency, timing, and length of time that the lake is open to the ocean may be less important for BFL than for SFL and YBF. The large annual fluctuations in BFL, SFL, and YBF are again typical of a fast growing, short-lived species in which abundance is very dependent on recruitment success into the lake. It

18 is likely that the BFL abundance may be controlled by environmental factors within the lake, whereas SFL and YBF abundance is dependent on factors such as lake opening, and conditions off the coast.

5. ACKNOWLEDGMENTS

This work was carried out by NIWA under contract to the Ministry of Fisheries (MFish project FLA200701. Thanks to Tony Adamson (Sanford Ltd) for providing the flatfish landing data, Dave Banks (SeaFIC) for the flatfish logbook data, Steve Brouwer (MFish) for manuscript review, Michael Stevenson (NIWA) for extracting trawl survey data, and Mike Beardsell (NIWA) for editorial comments.

6. REFERENCES

Beentjes, M.P. (2000). Assessment of red cod stocks for (RCO 3 and RCO 7) for 1999. New Zealand Fisheries Assessment Report 2000/25 . 78 p. Beentjes, M.P. (2003). Review of flatfish catch data and species composition. New Zealand Fisheries Assessment Report 17 . 22 p. Beentjes, M.P.; Renwick, J.A. (2001). The relationship between red cod, Pseudophycis bachus , recruitment and environmental variables in New Zealand. Environmental Biology of Fishes 61 : 315–328. Beentjes, M.P.; Stevenson, M.L. (2008). Inshore trawl survey of Canterbury Bight and Pegasus Bay, May-June 2007 (KAH0705). New Zealand Fisheries Assessment Report 2008/38 . 95 p. Blackwell, R. G.; Manning, M. J.; Gilbert, D. J. (2005). Standardised CPUE analysis of the target rig (Mustelus lenticulatus ) set net fishery in northern New Zealand (SPO 1 and SPO 8) Boyd, R.O.; Reilly, J.L. (2005). 1999/2000 national marine recreational fishing survey: harvest estimates. Final Research Report for Ministry of Fisheries project REC98/03. (Unpublished report held by MFish, Wellington.) Chambers, J. M.; Hastie, T. J. (1991). Statistical models in S. Wadsworth and Brooks/Cole, Pacific Grove. 608 p. Coburn, R.P.; Beentjes, M. P. (2005). Abundance estimates for flatfish in FLA 1 from standardised catch per unit effort analysis of the set net fisheries, 1989–90 to 2003–04. New Zealand Fisheries Assessment Report 2005/57 . 46 p. Colman, J. A. (1974). Movements of flounders in the Hauraki Gulf, New Zealand. New Zealand Journal of Marine & Freshwater Research 8: 79–93. Colman, J. A. (1978). Tagging experiments on the sand flounder, Rhombosolea plebeia (Richardson), in Canterbury, New Zealand, 1964 to 1966. Fisheries Research Bulletin 18 . 42 p. Colman, J. A. (1994). New Zealand flatfish. Seafood New Zealand . vol 2, issue 8 : 34–36. Dunn, A. (2002). Updated catch-per-unit-effort indices for hoki ( Macruronus novaezelandiae ) on the west coast South Island, Cook Strait, Chatham Rise, and sub-Antarctic for the years 1990 to 2001. New Zealand Fisheries Assessment Report 2002/47 . 51 p. Dunn, A.; Harley, S. J.; Doonan, I.; Bull, B (2000). Calculation and interpretation of catch-per-unit- effort (CPUE) indices. New Zealand Fisheries Assessment Report 2000/1 . 44 p. Kirk, P.D. (1989). Flatfish fishery assessment 1989. New Zealand Fisheries Assessment Research Document 89/8. 24 p. (Unpublished report held in NIWA library, Wellington.) Manning, M.J. (2007). Relative abundance of giant stargazer ( Kathetostoma giganteum ) in STA 5 based on commercial catch-per-unit-effort data. New Zealand Fisheries Assessment Report 2007/14 . 42 p. Manning, M.J.; Hanchet, S.M.; Stevenson, M.L. (2004). A description and analysis of New Zealand's spiny dogfish ( Squalus acanthias ) fisheries and recommendations on appropriate methods to monitor the status of the stocks. New Zealand Fisheries Assessment Report 2004/61 . 135 p. McCullagh, P.; Nelder, J.A. (1989). Generalised linear models. Chapman & Hall, London.

19 Ministry of Fisheries (2008a). Report from the Fishery Assessment Plenary, May 2008: stock assessments and yield estimates. Ministry of Fisheries. 990 p. (Unpublished report held in NIWA library, Wellington.) Ministry of Fisheries, Science Group (2008b). Guidelines to the design, implementation and reporting of catch sampling programmes. 16 p. (unpublished document held at Ministry of Fisheries, Wellington.) Ministry of Fisheries, Science Group (comps.) (2006). Report from the Fishery Assessment Plenary, May 2006: stock assessments and yield estimates. 875 p. (Unpublished report held in NIWA library, Wellington.) Parker, S.J.; Hurst, R.J.; Francis, R.I.C.C. (2009). Design options for an elephantfish, red gurnard, and flatfish complex trawl survey of the East Coast of the South Island. New Zealand Fisheries Assessment Report 2009/59 . 22 p. Roper, D.S.; Jillett, J.B. (1981). Seasonal occurrence and distribution of flatfish (Pisces: Pleuronectiformes) in inlets and shallow water along the Otago coast. New Zealand Journal of Marine & Freshwater Research 15 : 1–13. Starr, P J. (2007). Procedure for merging MFish landing and effort data, version 2.0. Report to the Adaptive Management Programme Fishery Assessment Working Group: Document 2007/04. 17 p. (Unpublished report held by Ministry of Fisheries, Wellington.) Stevens, D.W.; Francis, M.P.; Shearer, P.C.; McPhee, R.P.; Hickman, R.W.; Tait, M.J. (2005). Age and growth of two endemic flatfish ( Colistium guntheri and C. nudipinnis ) in central New Zealand waters. Marine and Freshwater Research 56 : 143–151. Stevens, D.W.; James, G.D.; Francis, M.P. (2004). Maximum age of New Zealand sole (Peltorhamphus novaezeelandiae ) from the west coast South Island. Final Research Report for Ministry of Fisheries Research Project FLA2003/01. 9 p. (Unpublished report held by Ministry of Fisheries, Wellington.) Stevenson, M.L.; Beentjes, M.P. (2001). Inshore trawl survey of the Canterbury Bight and Pegasus Bay, December 2000–January 2001 (KAH9917 & CMP9901). NIWA Technical Report 99 . 94 p.

20 Table 1: Product destination codes, greenweight and records for extracted flatfish data. Code that could potentially result in double-counting (shaded in grey) were dropped.

Code Greenweight (t) Records Description

L 32 372.773 139 154 Landed in New Zealand to a licensed fish receiver W 55.397 1 903 Sold at wharf U 20.781 1 706 Used as bait E 17.051 566 Eaten A 2.711 53 Accidental loss F 1.852 319 Recreational catch D 0.665 31 Discarded C 0.391 7 Disposed to the Crown S 0.001 1 Seized by the Crown

R 304.423 1 754 Retained on board Q 135.517 1 579 Holding receptacle on land T 45.916 82 Transferred to another vessel Null 7.935 64 Missing destination type code B 5.861 449 Stored as bait

32 971.274 147 668

Table 2: Application of the conversion factors (CF) to the various landed states using the most recent valid value specified for each product state code in the groomed and merged data.

Code Description Greenweight (t) Records CF type

GUT Gutted 32 445 141 158 GUT GRE Green (whole) 269 4 705 GRE HGU Headed and gutted 137 759 DRE Null Missing processed state code 52 32 – GGU Gilled and gutted 28 281 GUT DRE Dressed 14 147 DRE FIL Fillets:skin on 10 303 FIL MEA Fish meal 10 190 MEA Invalid Invalid processed state code 4 17 – SKF Fillets:skin off 1 55 SKF SUR Surimi 1 7 SUR DIS discarded 0 3 DIS EAT Eaten 0 5 EAT HGT Headed, gutted and tailed 0 3 DRE TRU Trunked 0 3 DRE

Totals 32 971 147 668

21 Table 3: Description of the CPUE model datasets used in the bottom trawl fishery analyses. Datasets are defined by method, target, statistical area, response variable, and catch score from the algorithm. The core vessels are defined by the minimum number of years present in the dataset and the minimum number of associated effort strata per year. Score 4: records include data from 1990–91 to 2006–07 and score “all” from 1989–90 to 2006–07. BT, bottom trawl; All, all target species (BAR, ELE, ESO, FLA, GUR, LSO, RCO, SFL, SPD, SPE, STA, TAR, & YBF).

Dataset definition Core vessel definition Model Method Target Stat area Response variable Catch score type Min. years Min. effort Cum. Prop. present strata per of catch year 0.1 BT All 020, 022–024, 026, 030 Log(FLAcatch) All 5 10 0.8 0.2 BT All 020, 022–024, 026, 030 Log(FLAcatch ) Score ‘4’ only 5 10 0.8 0.3 BT All 020, 022–024, 026, 030 Log(ESOcatch) All 5 10 0.8 0.4 BT All 020, 022–024, 026, 030 Log(ESOcatch) Score ‘4’ only 5 10 0.8 0.5 BT All 020, 022–024, 026, 030 Log(LSOcatch) All 5 10 0.8 0.6 BT All 020, 022–024, 026 , 030 Log(LSOcatch) Score ‘4’ only 5 10 0.8 0.7 BT All 020, 022–024, 026, 030 Log(SFLcatch) All 5 10 0.8 0.8 BT All 020, 022–024, 026, 030 Log(SFLcatch) Score ‘4’ only 5 10 0.8

Table 4: Description of the CPUE model datasets used in the setnet fishery analyses. Datasets are defined by method, target, statistical area, response variable, and catch score from the algorithm. The core vessels are defined by the minimum number of years present in the dataset and the minimum number of associated effort strata per year. Score 4: records include data from 1990–91 to 2006–07 and score “all” from 1989–90 to 2006–07. The datasets were further restricted by selecting only those effort strata where the associated landing name was “Lake Ellesmere”. Fishing years 1989–90 in model 0.11 (BFL), 1989–90 and 1993–94 in models 0.15 and 0.16 (YBF) have no associated non-zero catch records of each corresponding species; these years were dropped. SN, setnet.

Dataset definition Core vessel definition Model Method Target Stat area Response variable Catch score Min. years Min. effort Cum. Prop. type present strata per of catch year 0.9 SN BFL, FLA, SFL, YBF 022 Log(FLAcatch) All 3 10 0.8 0.10 SN BFL, FLA, SFL, YBF 022 Log(FLAcatch) Score ‘4’ only 3 10 0.8 0.11 SN BFL, FLA, SFL, YBF 022 Log(BFLcatch) All 3 10 0.8 0.12 SN BFL, FLA, SFL, YBF 022 Log(BFLcatch) Score ‘4’ only 3 10 0.8 0.13 SN BFL, FLA, SFL, YBF 022 Log(SFLcatch) All 3 10 0.8 0.14 SN BFL, FLA, SFL, YBF 022 Log(SFLcatch) Score ‘4’ only 3 10 0.8 0.15 SN BFL, FLA, SFL, YBF 022 Log(YBFcatch) All 3 10 0.8 0.16 SN BFL, FLA, SFL, YBF 022 Log(YBFcatch) Score ‘4’ only 3 10 0.8

Table 5: Percentage of flatfish species from 1) FLA 3 landed catch into Sanford (1992–93 to 2007–08, excluding 2001–02 and 2002–03); 2) Estimated catch from SeaFIC logbook scheme in southeast trawl fishery; 3) Biomass from ECSI Kaharoa winter and summer trawl surveys. FLA, flatfish; LSO, lemon sole; ESO, New Zealand sole; SFL, sand flounder; BFL, black flounder; BRI, Brill; YBF, yellowbelly flounder; GFL, greenback flounder.

Kaharoa surveys (mean % ) Sanford’s Logbook Winter Winter Summer Summer Species landings (%) catch (%) (30–400 m) (10–30 m) (30–400 m) (10–30 m)

FLA 3.7 0 0 0 0 BFL 0.0 0 0 0 0 BRI 5.6 4.2 1 0 0.6 2.0 ESO 32.8 61.1 10.3 30.9 21.4 38.9 GFL 0.02 0 1.2 0.0 11.0 11.0 LSO 38.8 11.0 80.2 4.6 63.5 4.8 SFL 15.5 22.6 7.2 49.0 3.6 37.6 TUR 0.1 0.3 0.2 11.2 0 0 YBF 0.05 0.9 0 4.3 0 5.7

22 FLA 1

FLA 2

FLA 2

FLA 7

FLA 3

FLA 3

Figure 1: Flatfish Quota Management Areas.

3000

2500

2000

1500

1000 FLA 3landings FLA (t) landings (FLA 3) 500 TACC

0

1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 Fishing year

Figure 2: Landings of flatfish in FLA 3 from 1983–04 to 2006–07 and the TACC (data from Ministry of Fisheries 2008a). 2004 = 2003–04 fishing year.

23

Figure 3: Map of Te Waihora (Lake Ellesmere). (Map reproduced by permission of Land Information New Zealand).

100 All QMAs FLA 1 FLA 2 FLA 3 FLA7 80

60

40 Percentcatch

20

0

0 2 3 4 6 7 /9 /91 /9 /9 /9 /9 /9 /00 1 2 5 6 9 89 90 93 00/01 9 9 99 99 9 99 99 0 1 1 1 1 1 1994/95 1 1 1997/98 1998/99 199 2 2001/02 Fishing year

Figure 4: The percentage of estimated flatfish catch reported by species code FLA for all QMAs combined, and for each QMA separately from 1989–90 to 2001–02 (from Beentjes 2003).

24

Figure 5: Algorithm (top panel) and the resultant flatfish catch by score (1-4) (bottom panel) used to estimate species specific catches. Scores 1 and 2 use the Sanford’s data and apply only to the trawl catch outside Te Waihora.

25

Figure 6: The relationship between key quantities in the groomed and merged datasets. The QMR catch (grey histogram), with the TACC (solid line), the landed greenweight catch (broken line with dots, from the warehou system; note the discrepancy with the QMR catch in the mid 1990s), and the total estimated catch as a proportion of the total landed greenweight catch per fishing year (broken line).

Figure 7: Using the groomed and merged flatfish datasets, the total estimated flatfish catch per fishing trip is plotted against the total landed flatfish catch per fishing trip on the left. Note the large number of zeros. Landed catch as a proportion of estimated catch is plotted as a histogram on the right. Note also the large number of zeros showing that many vessels have returned a landed greenweight record but no corresponding estimated catch in their corresponding effort records.

26

Figure 8: For all aggregated flatfish species codes, catches by year and month, statistical area, method, and target species.

27

Figure 9: For all aggregated flatfish species codes, catches by year and month. BT, bottom trawl; SN, setnet.

28

Figure 10: For all aggregated flatfish species codes, catches by year, method and from type. CEL, catch effort landing return; NCE, net catch effort landing return; TCP, trawl catch effort processing return.

29

Figure 11: For all aggregated flatfish species codes, catches by year, form type, target species, and method.

30

Figure 12: For all aggregated flatfish species codes, catches by year, form type, and statistical area and method.

31

Figure 13: For all aggregated flatfish species codes, catches by year, target species, statistical area, and method.

32

Figure 13 – continued

33

Figure 14: Inputted catch histories for individual flatfish species by year, and statistical area after applying the species catch estimation algorithm to the merged and groomed dataset

34

Figure 14 – continued

35

FLA FLA ESO ESO

LSO LSO SFL

SFL

Figure 15: Proportion of zero-catch records in each dataset (models 0.1 to 0.8) for the trawl CPUE models.

FLA Model 0.1 (all scores) FLA Model 0.2 (score 4 records only)

FLA FLA

Figure 16: Canonical indices for FLA CPUE models 0.1 and 0.2 (bottom trawl mixed target). The accepted models were - model 0.1: ln(Catch) ~ fishing year + P(effort number, 3) + target species + vessel key. Model 0.2: ln(Catch) ~ fishing year + P(fishing duration, 3) + vessel key + target species + stat.area.

36 ESO Model 0.3 (all scores) ESO Model 0.4 (score 4 records only)

ESO ESO

Figure 17: Canonical indices for ESO CPUE models 0.3 and 0.4 (bottom trawl mixed target). The accepted models were - model 0.3: ln(Catch) ~ fishing year + P(effort number, 3) + target species + vessel key + P(fishing duration, 3). Model 0.4: ln(Catch) ~ fishing year + P(fishing duration, 3) + vessel key + target species.

LSO Model 0.5 (all scores) LSO Model 0.6 (score 4 records only)

LS O LSO

Figure 18: Canonical indices for LSO CPUE models 0.5 and 0.6 (bottom trawl mixed target). The accepted models were - model 0.5: ln(Catch) ~ fishing year + P(effort number, 3) + target species + vessel key + month + stat.area. Model 0.6: ln(Catch) ~ fishing year + P(effort number, 3) + vessel key + month + stat.area + P(fishing duration, 3) + target species.

37 SFL Model 0.7 (all scores) SFL Model 0.8 (score 4 records only)

SFL SFL

Figure 19: Canonical indices for SFL CPUE models 0.7 and 0.8 (bottom trawl mixed target). The accepted models were - model 0.7: ln ln(Catch) ~ fishing year + target species + P(fishing duration, 3) + vessel key. Model 0.8: ln(Catch) ~ fishing year + P(fishing duration, 3) + vessel key + target species + stat.area + month.

FLA FLA BFL BFL

SFL YBF YBF

SFL

Figure 20: Proportion of zero-catch records in each dataset (models 0.1 to 0.8) for the setnet CPUE models.

38 FLA Model 0.9 (all scores) FLA Model 0.10 (score “4” records only)

Figure 21: Canonical indices for FLA CPUE models 0.9 and 0.10 for the Te Waihora setnet fishery. The accepted models were - model 0.9: ln(Catch) ~ fishing year + vessel key + P(total net set, 3) + month. Model 0.10: ln(Catch) ~ fishing year + vessel key + P(total net set, 3) + month.

BFL Model 0.11 (score “all” records) BFL Model 0.12 (score “4” records only)

Figure 22: Canonical indices for BFL CPUE models 0.12 for the Te Waihora setnet fishery. The accepted models were – model 0.11: ln(Catch) ~ fishing year + vessel key + month + target species + P(total net set, 3). Model 0.12: ln(Catch) ~ fishing year + vessel key + month + target species + P(total net set, 3).

39 SFL Model 0.13 (all scores) SFL Model 0.14 (score “4” records only)

Figure 23: Canonical indices for SFL CPUE models 0.13 and 0.14 for the Te Waihora setnet fishery. The accepted models were - model 0.13 ln(Catch) ~ fishing year + vessel key + target species + P(total net set, 3) + month + P(fishing duration, 3). Model 0.14: ln(Catch) ~ fishing year + vessel key + month + P(fishing duration, 3) + P(effort width, 3 ).

YBF Model 0.15 (all scores) YBF Model 0.16 (score “4” records only)

Figure 24: Canonical indices for YBF CPUE models 0.15 and 0.16 for the Te Waihora setnet fishery. The accepted models were - model 0.15 ln(Catch) ~ fishing year + vessel key + target species + month + P(total net set, 3) + P(effort width, 3). Model 0.16: ln(Catch) ~ fishing year + vessel key + target species + month + P(total net set, 3) + P(effort width, 3).

40

100 BRI 90 ESO 80 LSO 70 SFL FLA 60 50 40

30 20 Percent flatfish species catch 10 0

05 07 08 003 006 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2 2004 20 2 20 20 Fishing year

180 BRI 160 ESO 140 LSO 120 SFL FLA 100

80

60

Flatfish species catch (t) 40

20

0

3 4 5 6 7 1 2 3 4 5 6 7 8 0 0 0 0 00 199 199 199 199 199 1998 1999 2000 20 20 20 20 2 200 200 200 Fishing year

Figure 25: The percentage (top) and catch (t) (bottom) of the main flatfish species caught in FLA 3 (excluding Te Waihora) landed into Sanford Ltd. There are no data for 2002 and 2003. 1993 represents 1992–93 fishing year. BRI, brill; ESO, New Zealand sole; LSO, lemon sole; SFL, sand flounder; FLA, generic flatfish.

41 300 Lemon sole (LSO) N=2304 250

200

150

Numbers 100

50

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Total length (cm)

600 New Zealand sole (ESO) N=3853 500

400

300

Numbers 200

100

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Total length (cm)

300 Sand flounder (SFL) N=2753 250

200

150 ss Numbers 100

50

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Total length (cm)

Figure 26: Length frequency distributions from flatfish species caught by the commercial fishery in FMA 3. Data are from SeaFIC flatfish logbook programme for years 2004–05, 2005–06, and 2006–07 and are from 10 vessels.

42 100 Brill (BRI) N=557 80

60

40 Numbers

20

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Total length (cm)

20 Yellowbelly flounder (YBF) N=124

15

10 Numbers

5

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Total length (cm)

4 Turbot (TUR) N=12 3

2 Numbers

1

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Total length (cm)

Figure 26 –continued

43 100 ECSI w inter traw l surveys (30–400 m) 80

60

40

flatfish biomass flatfish 20 Mean proportion of of proportion Mean

0 BRI ESO GFL LSO SFL TUR YBF Species

ECSI 2007 w inter traw l survey 60 (10–30 m) 50 40 30 20

flatfish biomass flatfish 10 Mean proportion of of proportion Mean 0 BRI ESO GFL LSO SFL TUR YBF Species

ECSI summer traw l surveys 80 (30–400 m)

60

40

20 flatfish biomass flatfish Mean proportion of of proportion Mean 0 BRI ESO GFL LSO SFL TUR YBF Species

60 ECSI summer traw l surveys (10–30 m) 50 40 30 20 flatfish biomass flatfish

Mean proportion of of proportion Mean 10 0 BRI ESO GFL LSO SFL TUR YBF Species

Figure 27: The mean percentage of flatfish species biomass on the ECSI Kaharoa winter trawl surveys from 1991 to 2008 (N=7 surveys in 30–400 m and 1 survey in 10–30 m) and summer trawl surveys from 1996 to 2000 (N=5).. Error bars are 95% confidence intervals. BRI, brill; ESO, New Zealand sole; GFL, greenback flounder, LSO, lemon sole; SFL, sand flounder; TUR, turbot; YBF, yellowbelly flounder.

44

80 180 40 New Zealand sole (ESO) Lemon sole (LSO) Sand flounder (ESO) 70 160 35 winter (30–400 m) winter (30–400 m) winter (30–400 m) 60 140 30 120 50 25 100 40 20 80 30 15 60 20 40 10 10 20 5 0 0 0 1991 1992 1993 1994 1996 2007 2008 1991 1992 1993 1994 1996 2007 2008 1991 1992 1993 1994 1996 2007 2008 Survey year Survey year Survey year

180 New Zealand sole (ESO) 60 Lemon sole (LSO) 160 160 summer (10–30 m) 140 Sand flounder (SFL) 50 summer (10–30 m) 140 120 summer(10–30 m) 120 40 100 100 30 80 80 60 60 20 40 40 10 20 20 0 0 0 1996 1997 1998 1999 2000 1996 1997 1998 1999 2000 1996 1997 1998 1999 2000 Survey year Survey year Survey year

300 350 60 New Zealand sole (ESO) Lemon sole (LSO) Sand flounder (SFL) 300 250 summer (30–400 m) summer (30–400 m) 50 summer (30–400 m) 250 200 40 200 150 30 150 100 20 100 50 50 10 0 0 0 1996 1997 1998 1999 2000 1996 1997 1998 1999 2000 1996 1997 1998 1999 2000 Survey year Survey year Survey year

Figure 28: Kaharoa ECSI survey biomass indices for winter (30–400 m) and summer surveys (10–30 m and 30–400 m) for ESO, LSO, and SFL. Error bars are 95% confidence intervals.

45

Winter ECSI Kaharoa surveys

12 Lemon sole (LSO) m=295 10 f=1327 unsexed 8 u=761 female 6 male

Percent 4

2

0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

10 New Zealand sole (ESO) m=45 f=106 8 unsexed u=89 6 females males 4 Percent

2

0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

9 Sand flounder (SFL) 8 m=45 7 f=132 unsexed u=10 6 female 5 male 4 Percent 3 2 1 0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

Figure 29: Scaled length frequency distributions of lemon sole, New Zealand sole, and sand flounder from ECSI Kaharoa winter trawl surveys from 1992 to 2008 in 30–400 m (N=6, kah0705 includes 10–30 m). Length frequency data were summed across all surveys. m, number of males measured; f, number of females measured; u, number of unsexed fish measured.

46 Summer ECSI Kaharoa surveys

10 Lemon sole (LSO) 9 m=101 8 f=213 unsexed 7 u=2738 6 females 5 males 4 Percent 3 2 1 0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

9 New Zealand sole (ESO) 8 m=144 7 f=223 unsexed u=2547 6 females 5 males 4 Percent 3 2 1 0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

12 Sand flounder (SFL) m=10 10 f=48 unsexed u=985 8 females 6 males

Percent 4

2

0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

Figure 30: Scaled length frequency distributions of lemon sole, New Zealand sole, and sand flounder from ECSI Kaharoa summer trawl surveys from 1996 to 2000 in 10–400 m (N=5). Length frequency data were summed across all surveys. m, number of males measured; f, number of females measured; u, number of unsexed fish measured.

47 Summer ECSI Kaharoa surveys

16 Greenback flounder (GFL) unsexed 14 m=0 females 12 f=6 males 10 u=222 8

Percent 6 4 2 0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

30 Yellowbelly flounder (YBF) 25 m=4 f=1 unsexed 20 u=117 females 15 males

Percent 10

5

0 0 5 10 15 20 25 30 35 40 45 50 55 60 Total length (cm)

Figure 30 – continued

48

KAH9105 KAH9205

Waiau River Waiau River

   43° S   43° S                      Banks  Banks   Peninsula   Peninsula                44°    44°                                                                                 45°  45°       Shag  Shag Point  Point     171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9306 KAH9406

Waiau River Waiau River

   43° S   43° S              Banks  Banks       Peninsula      Peninsula                  44°        44°                                                                            45°    45°             Shag   Shag        Point  Point       

171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9606

Waiau River

43° S             Banks  Peninsula                44°                                                                     45°     Shag      Point  

171° E 172° 173° 174°

Figure 31: Catch rates for New Zealand sole, from the ECSI Kaharoa winter trawl survey series (30–400 m). (maximum catch rate = 32.1 kg/km 2). Open circles = no catch. Solid circles are proportional to the catch rate.

49

KAH0705 KAH0806

Waiau River Waiau River

  43° S  43° S                               Banks    Banks       Peninsula  Peninsula                      44°   44°                                                                                                                     45°  45°           Shag   Shag         Point  Point 

171° E 172° 173° 174° 171° E 172° 173° 174°

Figure 31–continued (ESO).

50 KAH9618 KAH9704

Waiau River Waiau River

  43° S   43° S                                                           Banks  Banks       Peninsula  Peninsula                                         44°    44°                                                                                                                                                     45°   45°                    Shag  Shag  Point  Point      171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9809 KAH9917

Waiau River Waiau River

 43° S  43° S                                          Banks     Banks    Peninsula     Peninsula                                      44°                    44°                                                                                                                                45°        45°    Shag  Shag          Point  Point       171° E 172° 173° 174° 171° E 172° 173° 174°

KAH0014

Waiau River

  43° S              Banks  

Peninsula                       44°                                                                      45°  Shag    Point   

171° E 172° 173° 174°

Figure 32: Catch rates for New Zealand sole (ESO) from the ECSI Kaharoa summer trawl survey series (10–400 m). (Maximum catch rate = 183 kg/km 2). Open circles = no catch. Solid circles are proportional to the catch rate.

51

KAH9105 KAH9205

Waiau River Waiau River

 

 43° S   43° S                           Banks  Banks      Peninsula   Peninsula                    44°   44°                                                                                                   45°  45°         Shag Shag 

Point  Point       171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9306 KAH9406

Waiau River Waiau River

   43° S 43° S                    Banks   Banks            Peninsula   Peninsula                                 44°                  44°                                                                                              

            45°     45°              Shag  Shag                   Point   Point              171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9606

Waiau River

43° S                Banks 

Peninsula                    44°                                                                         45°       Shag        Point  

171° E 172° 173° 174°

Figure 33: Catch rates for lemon sole from the ECSI Kaharoa winter trawl survey series (30–400 m). (Maximum catch rate = 93.4 kg/km2). Open circles = no catch, solid circles are proportional to the catch rate.

52

KAH0705 KAH0806

Waiau River Waiau River

  43° S  43° S                                   Banks    Banks      Peninsula Peninsula                          44°   44°                                                                                                                                45° 45°                   Shag  Shag        Point  Point  171° E 172° 173° 174° 171° E 172° 173° 174°

Figure 33 – continued (LSO).

53

KAH9618 KAH9704

Waiau River Waiau River

  43° S   43° S                                                                      Banks   Banks        Peninsula  Peninsula                                               44°    44°                                                                                                                                                                        45°  45°                   Shag   Shag   Point  Point         171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9809 KAH9917

Waiau River Waiau River

  43° S   43° S                                   Banks    Banks           Peninsula      Peninsula                                       44°            44°                                                                                                                                                  45°             45°        Shag Shag               Point   Point             171° E 172° 173° 174° 171° E 172° 173° 174°

KAH0014

Waiau River

  43° S            Banks  

Peninsula                       44°                                                                       45°

 Shag     Point   

171° E 172° 173° 174° Figure 34: Catch rates for lemon sole from the ECSI Kaharoa summer trawl survey series (10–400 m). (Maximum catch rate = 134 kg/km 2). Open circles = no catch. Solid circles are proportional to the catch rate.

54 KAH9105 KAH9205

Waiau River Waiau River

   43° S    43° S                     Banks  Banks    Peninsula   Peninsula               44°   44°                                                                                    45°  45°      

Shag  Shag Point  Point     171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9306 KAH9406

Waiau River Waiau River

   43° S 43° S                  Banks   Banks         Peninsula  Peninsula                         44°        44°                                                                                       45°    45°        Shag   Shag         Point  Point         171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9606

Waiau River

43° S             Banks  

Peninsula                  44°                                                                      45°     Shag      Point  

171° E 172° 173° 174°

Figure 35: Catch rates for sand flounder from the ECSI Kaharoa winter trawl survey series (30–400 m). (Maximum catch rate = 67.8 kg/km 2). Open circles = no catch. Solid circles are proportional to the catch rate.

55

KAH0705 KAH0806

Waiau River Waiau River

  43° S  43° S                                   Banks    Banks      Peninsula Peninsula                          44°   44°                                                                                                                                45° 45°                   Shag  Shag        Point  Point  171° E 172° 173° 174° 171° E 172° 173° 174°

Figure 35–continued (SFL).

56

KAH9618 KAH9704

Waiau River Waiau River

  43° S   43° S                                                           Banks  Banks    Peninsula  Peninsula                                         44°   44°                                                                                                                                             45°   45°                 Shag   Shag  Point  Point      171° E 172° 173° 174° 171° E 172° 173° 174°

KAH9809 KAH9917

Waiau River Waiau River

 43° S  43° S                                    Banks     Banks       Peninsula     Peninsula                                             44°                44°                                                                                                                               45°        45°     Shag  Shag           Point  Point       171° E 172° 173° 174° 171° E 172° 173° 174°

KAH0014

Waiau River

  43° S           Banks  

Peninsula                         44°                                                                       45°  Shag    Point   

171° E 172° 173° 174°

Figure 36: Catch rates for sand flounder from the ECSI Kaharoa summer trawl survey series (10–400 m). (Maximum catch rate = 178 kg/km 2). Open circles = no catch. Solid circles are proportional to the catch rate.

57 3000 1.6

1.4 2500 1.2 2000 1

1500 0.8

0.6 1000 FLA 3 (t) landings

FLA 3 landings CPIE Standardised 0.4 FLA model 0.1 (all scores) 500 FLA model 0.2 (score 4 only) 0.2

0 0

8 94 00 06 85–86 87–8 91–92 93– 97–98 99– 03–04 05– 1983–84 19 19 1989–90 19 19 1995–96 19 19 2001–02 20 20 Fishing year

Figure 37: Landed FLA 3 catch from 1983–04 to 2006–07 (data from Ministry of Fisheries 2008a) and standardised CPUE indices for the trawl fishery aggregated flatfish model 0.1 (all scores) and model 0.2 (score 4 only).

200 800 All species (ESO) All species winter (30–400 m) 150 600 summer (10–400 m)

100 400

50 200

0 0 1996 1997 1998 1999 2000 Survey year Survey year

Figure 38: Kaharoa biomass of all flatfish species for winter and summer surveys.

500 All species 400

300 200

100

Flatfish species Flatfish catch(t) 0

3 7 5 01 199 1995 199 1999 20 2003 200 2007 Fishing year

Figure 39: Sanford flatfish landings data for all flatfish catch combined.

58 Appendix 1. Expected catch rates (log catch), model diagnostic plots, and fleet history plots for models 0.1 to 0.8 (bottom trawl mixed target fishery).

FLA Model 0.1 (all scores)

59 FLA Model 0.1 (all scores)

FLA Model 0.2 (score “4” records only)

60 FLA Model 0.2 (score “4” records only)

61 ESO Model 0.3 (all scores)

62 ESO Model 0.3 (all scores)

ESO Model 0.4 (score “4” records only)

63 ESO Model 0.4 (score “4” records only)

64 LSO Model 0.5 (all scores)

65 LSO Model 0.5 (all scores)

LSO Model 0.6 (score “4” records only)

66 LSO Model 0.6 (score “4” records only)

67 SFL Model 0.7 (all scores)

68 SFL Model 0.7 (all scores)

SFL Model 0.8 (score “4” only)

69 SFL Model 0.8 (score “4” only)

70 FLA Model 0.9 (all scores)

71 FLA Model 0.9 (all scores)

FLA Model 0.10 (score “4” only)

72 FLA Model 0.10 (score “4” only)

73 BFL Model 0.11 (all scores)

74 BFL Model 0.11 (all scores)

BFL Model 0.12 (score “4” only)

75 BFL Model 0.12 (score “4” only)

76 SFL Model 0.13 (all scores)

77 SFL Model 0.13 (all scores)

SFL Model 0.14 (score “4” only)

78 SFL Model 0.14 (score “4” only)

79 YBF Model 0.15 (all scores)

80 YBF Model 0.15 (all scores)

YBF Model 0.16 (score “4” only)

81 YBF Model 0.16 (score “4” only)

82