Stock Assessment Form Sprat

Transcription

1 Stock Assessment Form Sprat Reporting year: 2017 Reference year: 2016 Sprat represent a unit stock shared among the Black Sea countries. Its key role is determined by the importance from both commercial and ecological point of view. The sprat fishery takes place in the Black Sea (GFCM Fishing Sub-area 37.4 (Division ) and Geographical Sub-area (GSA) 29). The sprat landings highly varied as for the average catch account ed to tons. In 2014 the total catch accounted tons. The total catch reached t and t in 2015 and 2016, respectively. The Turkish fleet takes the largest part of the catches ( t and t). We was used ICA Model. In 2016, SSB is estimated at t, which is one of the highest estimated of the time series. Recruitment has been low in but has been increasing since The current explotation rate (E = 0.36, which corresponds to an F = 0.54) is smaller than EMSY (0.40, which corresponds to an F = 0.64), indicatings that sprat in GSA 29 is being fished below EMSY.

2 Stock Assessment Form version 0.9 Uploader: Salih İLHAN Stock assessment form Pelagic Sprat GSA29 Sommario 1 Basic Identification Data Stock identification and biological information Stock unit Growth and maturity Fisheries information Description of the fleet Historical trends Commercial CPUE Management regulations Reference points Fisheries independent information Pelagic survey Stock Assessment Integrated catch-at-age analysis (ICA) Model assumptions Scripts Input data and Parameters Results Model assumptions Scripts Input data and Parameters Results Retrospective analysis, comparison between model runs, sensitivity analysis, etc Assessment quality Stock predictions Draft scientific advice Explanation of codes... 52

3 1 Basic Identification Data Scientific name: Common name: ISCAAP Group: Sprattus sprattus L. Sprat 37 1 st Geographical sub-area: 2 nd Geographical sub-area: 3 rd Geographical sub-area: 29 Bulgaria Romania Ukraine Russian Federation Turkey Georgia Stock assessment method: (direct, indirect, combined, none) Indirect for GSA29 *Salih İLHAN, Erdal ÜSTÜNDAĞ **Daskalov, G.; Raykov, V.; Georgieva, Y. Authors: Affiliation: *Central Fisheries Research Instıtute, TURKEY **Institute of Oceanology - BAS, IBER-BAS The ISSCAAP code is assigned according to the FAO 'International Standard Statistical Classification for Aquatic Animals and Plants' (ISSCAAP) which divides commercial species into 50 groups on the basis of their taxonomic, ecological and economic characteristics. This can be provided by the GFCM secretariat if needed. A list of groups can be found here: Direct methods (you can choose more than one): - Acoustics survey - Egg production survey - Trawl survey Indirect method (you can choose more than one): - ICA -X - VPA - LCA - AMCI - XSA - Biomass models - Length based models - Other (please specify) Combined method: you can choose both a direct and an indirect method and the name of the combined method (if it does exist).

4 2 Stock identification and biological information The Black Sea sprat (Sprattus sprattus L.) is a key species in the Black Sea ecosystem. Sprat is a marine pelagic schooling species, sometimes entering in the estuaries (especially as juveniles) and the Azov Sea and tolerating salinities as low as 4. Sprat is one of the most important fish species, being fished and consumed traditionally in the Black Sea countries. It is most abundant small pelagic fish species in the region, together with anchovy and horse mackerel and accounts for most of the landings in the north-western part of the Black Sea. Whiting is also taken as a by-catch in the sprat fishery, although there is no targeted fishery beyond this (Raykov, 2006) except for Turkish waters. Sprat fishing takes place on the continental shelf on m of depth (Shlyakhov, Shlyakhova, 2011). The harvesting of the Black Sea sprat is conducted during the day time when its aggregations become denser and are successfully fished with trawls. The main fishing gears are mid-water otter trawl, pelagic pair trawls and uncovered pound nets. The species is fast growing; age comprises 4-5 age groups. Sprat has lengths comprised between 50 and 120 mm, the highest frequency pertaining to the individuals of mm lengths. The age corresponding to these lengths was , the ages having a significant participation. By 1982, the age classes 4-4+ years had a share of 34% from the catch of this species, then the percentage continually decreased up to 1995 when this age was not signalled, meaning the increase of the pressure through fishing exerted on the populations. While the share of this age decreased, the prevalence of 0+ especially 1-1+ ages became increased. During last years the age structure show the presence of the specimens of and 3; 3 + years, the catch base being the individuals of and years. The sprat fishery is taking place in the Black Sea (GFCM Fishing Sub-area 37.4 (Division ) and Geographical Sub-area (GSA) 29). The opportunities of marine fishing are limited by the specific characteristics of the Black Sea. The exploitation of the fish recourses is limited in the shelf area. The water below m is anoxic and contains hydrogen sulphide. In Bulgarian, Romanian, Russian and Ukrainian waters the most intensive fisheries of Black Sea sprat is conducted in April till October with mid-water trawls on vessels m long and a small number vessels >40m. Beyond the 12-mile zone a special permission is needed for fishing. Harvesting of Black Sea sprat is conducted during the day, when the sprat aggregations become denser and are successfully fished with mid-water trawls. The significance of the sprat fishery in Turkey in the last three years has increased and the landings reached 77 thous.t. t in The landings are 50 thous. t in The main gears used for sprat fishery in Turkey (fishing area is constrained in front of the city of Samsun) are pelagic pair trawls working in spring at m depth and in autumn - in deeper water: m depths. 2.1 Stock unit It is assumed that sprat represent a unit stock shared among the Black Sea countries 2.2 Growth and maturity No maturity studies carried out in All fish (100%) are assumed to mature at the end of the first year of their life.

5 Table 2.2-1:Maximum size, size at first maturity and size at recruitment. Somatic magnitude measured (LH, LC, etc)* Units* Sex Fem Mal Both Unsexed Maximum size observed 13 Reprod uction seas on Size at first maturity 6 Reprod uction as are Nov-March North western Black Sea Recruitment size Nursery areas North western Black Sea coastal zone and marginal habitats

6 sex ratio (% females/total)

7 3 Fisheries information The sprat fishery is taking place in the Black Sea (GFCM Fishing Sub-area 37.4 (Division ) and Geographical Sub-area (GSA) 29). The opportunities of marine fishing are limited by the specific characteristics of the Black Sea. The exploitation of the fish recourses is limited in the shelf area. The water below m is anoxic and contains hydrogen sulphide. In Bulgarian, Romanian, Russian and Ukrainian waters the most intensive fisheries of Black Sea sprat is conducted in April till October with mid-water trawls on vessels m long and a small number vessels > 40m. Beyond the 12-mile zone a special permission is needed for fishing. Harvesting of Black Sea sprat is conducted during the day, when the sprat aggregations become denser and are successfully fished with mid-water trawls. The significance of the sprat fishery in Turkey in years has increased and the landings reached 77 thous. t in In 2016, Turkey catched 50 thous t. The main gears used for sprat fishery in Turkey (fishing area is constrained in front of the city of Samsun) are pelagic pair trawls working in spring at m depth and in autumn - in deeper water: m depths.

8 3.1 Description of the fleet Table 3.1-1: Description of operational units in the stock Country GS A Fleet Segmen t Fishin g Gear Class Group of Target Species Species Operationa l Unit 1 Bulgaria 29 24<40 12<18 18<24 6<12 OTM Sprat, horse mackerel,bluefish,anchovy Alosa immaculata,atherina pontica,raja clavata, Dasyatis pastinavca,m.merlangius, Squalus acanthias etc Operationa l Unit 2 Operationa l Unit 3 Operationa l Unit 4 Operationa l Unit 5 Bulgaria 29 - FPN GNS Sprat, anchovy, horse mackerel Romania 29 24<40 OTM Sprat, anchovy, horse mackerel Romania 29 - Ukraine 29 24<40 12<18 18<24 6<12 FPN,GN S Sprat, anchovy,horse mackerel Sprat, anchovy,horse mackerel Alosa immaculata,atherina pontica,raja clavata, Dasyatis pastinavca,m.merlangius,squal us acanthias etc Alosa immaculate,atherina pontica,raja clavata, Dasyatis pastinavca,m.merlangius, Squalus acanthias etc Alosa immaculata,atherina pontica,raja clavata, Dasyatis pastinavca,m.merlangius, Squalus acanthias etc Alosa immaculata,atherina pontica,raja clavata, Dasyatis pastinavca,m.merlangius, Squalus acanthias etc Operationa l Unit 6 Turkey 29 24<40 12<18 18<24 6<12 OTM, Pair trawls, Purse seiners Sprat, horse mackerel,bluefish,anchovy,boni to Alosa immaculata,atherina pontica,raja clavata, Dasyatis pastinavca,m.merlangius, Squalus acanthias etc Operationa l Unit 7 Russian Federatio n 29 24<40 12<18 18<24 6<12 OTM Sprat, horse mackerel,bluefish,anchovy Alosa immaculata,atherina pontica,raja clavata, Dasyatis pastinavca,m.merlangius, Squalus acanthias etc

9 Table 3.1-2: Catch, bycatch, discards and effort by operational unit Operational Units* Total Fleet (n of boats)* stationary nets and 2000 GNS 24 Kilos or Tons Catch (speci es assess ed) sprat sprat sprat 22 sprat sprat sprat sprat Other species caught M.merlangius,D.pastin aca,raja clavata,sq.acanthias,al osa immaculata, etc M.merlangius,D.pastin aca,raja clavata,sq.acanthias,al osa immaculata,atherina pontica etc M.merlangius,D.pastin aca,raja clavata,sq.acanthias,al osa immaculata,atherina pontica etc M.merlangius,D.pastin aca,raja clavata,sq.acanthias,al osa immaculata,atherina pontica etc M.merlangius,D.pastin aca,raja clavata,sq.acanthias,al osa immaculata,atherina pontica etc M.merlangius,D.pastin aca,raja clavata,sq.acanthias,al osa immaculata,atherina pontica etc M.merlangius,D.pastin aca,raja clavata,sq.acanthias,al osa immaculata,atherina pontica etc Discards (species assessed ) Discards (other species caught) no - no - no Effort units Kw*days/GT* days Days deployed Kw*days, GT*days Days, hours deployed Kw*days, GT*days Kw*days,GT* days Kw*days,GT* days

10 Table 3.1-3: Catches as used in the assessment Classification Catch (tn) BG 2295 GE - RO UKR TU RU Total Historical trends The following Tables list the fishing effort data received from Member States through the official DCF data call in units of kw*days at sea and number of vessels. As submitted to JRC through the DCF 2015 Med and Black Sea data call by major gear type 2015 in Bulgaria (NAFA,2015), 76% of the total sprat landings in Bulgarian marine area were realized by fleet segment 24<40 m LOA. In Romania only one fishing vessel using OTM targeting sprat has been operating in Black Sea. Major fishing gears used for sprat fishery were stationary uncovered pound nets.

11 3.3 Commercial CPUE Commercial CPUE kg.h -1 increased in 2015 in comparison with 2014 in Bulgarian part. The same trend is detected for the in Turkey sprat fishery but Commercial CPUE kg.h -1 decreased in 2016 in comparison with In Romanian waters a significant drop of CPUE has been observed due to drastic reduction of the fishing fleet. Table Effort of vessels using OTM and FPO targeting sprat in 2015, Bulgaria (NAFA,2015) initial year Description: OTM Description:FPO N vessels kw*days GT*days Hrs fished N vessels kw*days GT*days Hrs fished The main fishing gears targeting sprat in Bulgaria are OTM, FPO and BS. The distribution of CPUE to the corresponding fishing fleet segments are presented on Table.

12 Table Data regarding sprat fishery fleet in TURKEY, total landings and CPUE (TUIK Fishery Statistics) Years Total landing (tons) No of vessels CPUE (tons/years/vessel) , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,932

13 Table CPUE kg/h *1000 of Ukrainian fishing vessels (Shlyakhov et al );for Odessa Center of YugNIRO. Ukrainian commercial fleet CPUE kg*h-l by years and quarters Jan-Mar Apr-Jun Jul-Sep Oct-Dec Average na na na na na 2014* * * 900 *North-Westem part of Black Sea 3.4 Management regulations A quota is allocated in EU waters of the Black Sea (Bulgaria and Romania). No fishery management agreement exists among other Black Sea countries. In the EU Black Sea waters a global (both Romania and Bulgaria) TAC tons has been allocated in 2009 and In 2011 and in 2012 allocated quota in Bulgarian waters was at the rate of t sprat (Council Regulation 5/2012) and t for Romanian waters.the decreasing trend in indices since 2008 was observed despite of quotas regime in force in community waters. From the catches of fish only the turbot species (Scophthalmus maximus) and sprat (Sprattus sprattus) are subject to quotas and are included in the National data collection program (NDCP). The applied quotas are precautionary because it is not possible their biomass to be calculated for the whole water basin of the Black Sea.

14 Table EC quota and recommended Total allowable catch of sprat in EU waters for Quota according to Regulation (EU) 1579/2007. Regulation (EU) 1139/2008.Regulation (EU) 1287/2009.Regulation (EU) 1004/2010.Regulation (EU) 1256/2010. Regulation (EU) 5/ EC s quota 3. Source of data: Institute of Oceanology BAS. Bulgaria 4. Source of data: Institute of Oceanology BAS. Bulgaria and NIMRD,Romania 5. National Institute for Marine Research and Development, Romania Current management regulations are in force for the sprat fisheries in Turkey: (1) Regulations about fishing area: Sprat fishery by pelagic trawls should be conducted only along Samsun shelf area. The coordinates of this area were specified. But except sprat the fishery was allowed for anchovy, horse mackerel and bluefish along other trawling areas in Black Sea. (2) Regulations about fishing gear: In Turkey pelagic trawls operate as paired vessels. Vessels engaged in sprat fishery need to receive licence eligible only for one fishing period from Samsun City Directorate of Food. Agriculture and Livestock. The single vessel operation in pelagic fishery seems to be inconvenient for Turkey at least for now as the fisherman can quickly change the gear to bottom trawling during operation. (3) Regulations about time periods: Though pelagic fishing period starts in 15 September as same as bottom trawling. it lasts to 15 May. Bottom trawling ends with 15 April. There is no limitation in distance from land for pelagic trawling. (4) Regulations about depth: The pelagic fishery is banned in waters shallower than 18 m in fishing area between 15 September and 15 April. But between 15 April 15 May it is allowed in waters deeper than 36 m limited with offshore of Çayağzı Cape (Samsun-Yakakent) in west and Akçay estuary (Samsun Ordu city border) in east (Anonymous, 2006). Sprat catch reaches a maximum in this one month-period and provide a great economic input for fishermen. Conversely with bottom trawling depth limitations are in force in pelagic fishery instead distance from land. But as mentioned above the depth limitation is increased to 36 m by 15 April in order to protect spawning adults and juveniles on coastal zone.

15 Current management regulations are in force for the sprat fisheries in Russian Federation: (1) Regulations about fishing area: sprat fishery by mid-water trawls (OTM) should be conducted along areas of the Crimea and the Krasnodar Region, with the exception of a number of prohibited areas specified in the Regulations for the Azov-Black Sea Basin; during the year in the sea (in Karkinitsky Bay from September 1 to July 10) with trap nets and beach seines. (2) Regulations on fishing gear and time periods: in all areas of the sprat fishery min mesh size of trawls is 12 mm; trawling of sprats is allowed from April 1 to October 31 - in the sea to the west of Cape Meganom meridian (with the exception of Karkinitsky Bay) by mid-water trawls (size of trawls are not limited) and from April 1-25 (depending on the area) to October 31 - in the sea to the east of Cape Meganom meridian (the size of trawls is limited to 28 m and a displacement of fishing vessels is limited to 1300 register tons). Table Sprat total TAC (t) applied to vessels of Ukraine. Year Ukraine *) *) *) *) A TAC was not set in Ukraine in Table Minimum landing size of sprat in GSA 29. Legend: TL total length; SL standard length. BG GE RO RU TR UA Sprattus sparttus TL=7cm SL=6cm TL=7cm SL= 6cm NO SL=6cm 3.5 Reference points Table 3.5.1: List of reference points Criterion Current value Units Reference Point Trend Comments B SSB t increasing F 0.54 =F 0.64 decreasing Y CPUE

16 4 Fisheries independent information 4.1 Pelagic survey Survey -1 Bulgarian hydro acoustic survey Pelagic Trawl Survey was accomplished in August September and December 2016 in the Bulgarian Black Sea area. To establish the abundance of the reference species (Sprattus sprattus) in front of the Bulgarian coast a standard methodology for stratified sampling was employed (Gulland, 1966;). To address the research objectives the region was divided in four strata according to depth Stratum 1 (15 35 m) Stratum 2 (35 50 m), Stratum 3 (50 75 m) and Stratum 4 ( m). The study area in Bulgarian waters was partitioned into 128 equal in size not overlying fields, situated at depth between m. The total surveyed area in Bulgarian part was km -2 and total estimated biomass was t in August-September The total surveyed area in Bulgarian part in December was km -2 and total estimated biomass was t. The estimated abundance indices, CPUA (catch per unit area, kg/km 2 ) and the relative sprat biomasses (kg) during the Bulgarian Black Sea scientific survey (2016). Survey -2 Romanian mid-water trawl survey Pursuant to the Bilateral Agreement, Romania performed the pelagic surveys of 2016 in the Romanian Black Sea area, using the Steaua de Mare R/V, during quarters 2 (June) and 4 (Octomber - November). The following parameters were considered during the surveys: Mid water trawl (57/63-62 m): trawling speed ,5 Kts; horizontal opening - 22 m; trawling time - 30 min. The results obtained were presented as maps and tables comprising data on: surface of the swept grid (Nm 2, m 2 ); mean weight per area unit (g/m 2, t/mm 2 ); weight variation ranges per area unit; total biomass values (t). The survey was conducted at depths between 13.8 m and 62 m and covered almost entirely the continental shelf of the Romanian coast, between St. Gheorghe and Vama Veche. For this purpuse 42 hauls were performed in quarter 2, and 34 hauls in quarter 4. The distribution of the sprat agglomerations are shown on Figure 4.1.1

17 Figure Distribution of the sprat agglomerations in the 2nd Quarter 2016 (left) and 4nd Quarter 2016 (right) in Romanian marine waters. Table Assessment of sprat agglomerations in June 2015, pelagic trawl survey, Romanian area Depth range (m) 0-30m 30 50m m Total Investigated area (Nm 2 ) Variation of the catches (t/ Nm 2 ) Average catch (t/ Nm 2 ) Biomass of the fishing agglomerations (t) Biomass extrapolated the Romanian shelf (t) Table Assessment of sprat agglomerations in June 2016, pelagic trawl survey, Romanian area Depth range (m) 0-30m 30 50m m Total Investigated area (Nm 2 ) Variation of the catches (t/ Nm 2 ) Average catch (t/ Nm 2 ) Biomass of the fishing agglomerations (t) Biomass extrapolated the Romanian shelf (t)

18 Table Assessment of sprat agglomerations in November 2016, pelagic trawl survey, Romanian area Depth range (m) 0-30m 30 50m m Total Investigated area (Nm 2 ) Variation of the catches (t/ Nm 2 ) Average catch (t/ Nm 2 ) Biomass of the fishing agglomerations (t) Biomass extrapolated the Romanian shelf (t) Age composition of catches indicates the presence of 1-3 years old individuals. Most of the individuals caught are 1 years old (58.7%), followed by those of 2 years (32.4%) and 3 years (8.9%) (Figure ). 3 years 8.9 % 1 year 58.7 % 2 years 32.4 % TOTAL 2016 Sprat no Figure Structure by age composition of sprat as estimated by the Romanian trawl survey, References: Raykov V., Yankova M., Ivanova, P., Mihneva V., Dimitrov D., Trayanova A., Kotsev I., Djembekova N., Bekova R., Valcheva N., Pelagic trawl surveys Project report for the National Agency of Fisheries and Aquaculture of Bulgaria. National Data Collection program for 2016, 100 pp.

19 5 Stock Assessment 5.1 Integrated catch-at-age analysis (ICA) Model assumptions Catch-at-age Analysis (ICA; Patterson and Melvin. 1996) was used to assess the stock of sprat in GSA 29. ICA is a statistical catch-at-age method based on the Fournier and Deriso models (Deriso et al., 1985). It applies a statistical optimization procedure to calculate population numbers and fishing mortality coefficients-at-age from data of catch numbers-at-age and natural mortality. The dynamics of a cohort (generation) in the stock are expressed by two non-linear equations referred to as a survival equation (exponential decay) and a catch equation: Na+1.y+1 = Na.y*exp( Fa.y M). C a.y = N a.y *[1 exp( F a.y M)]* F a.y / (F a.y + M) where C, N, M and F are catch, abundance, natural mortality, and fishing mortality, while a and y are subscript indices for age and year. The algorithm initially estimates population numbers and fishing mortality fitting a separable model. when F is assumed to conform to a constant selection pattern (fishing mortality-at-age). but fishing mortality by year is allowed to vary. The F matrix is then modelled as a multiplication of the year-specific F and the specified selection pattern. This procedure substantially diminishes the number of parameters in the model. In its second stage. the ICA algorithm minimizes the weighted Sum of Square Residuals (SSR) of observed and modelled catch and relative abundance indices (CPUE). assuming Gaussian distribution of the log residuals: min [ a.y pca.y (log Ca.y log Ĉa.y)2 + a.y.f pia.f (log Ia.y.f log Î a.y.f)2 where C, Ĉ, I, and Î are observed and estimated catch and age-structured index, respectively, and a, y, and f are subscript indices for age, year and fleet. Weights associated with catches and different indices (pc, pi) are ideally set equal to the inverse variances of catch and index data and can be calculated based on the residuals between modelled and observed values. However. weights are usually set by the user on the basis of some information about the reliability of different indices and current experience with modelling the stock. Indices are defined as related to population numbers by the equations: Î a.y = Na.y*exp( Fa.y M) Î a.y = qa*na.y*exp( Fa.y M) Î a.y = qa*(na.y*exp( Fa.y M))ka The two unknown parameters (qa. an age-specific catchability. and k. a constant) are estimated according to the assumed relationship between the population and the abundance index, which has to be specified as being one of the above identity. linear or power, respectively. ICA combines the power and accuracy of a statistical model with the flexibility of setting different options of the parameters (e.g. a separable model accounting for age effects) and for this raison is suitable for a short living species (age 5 at maximum) such as the Black Sea sprat. ICA has previously been applied to Black Sea sprat by Daskalov (1998) and Daskalov et al. 2010, 2011, and 2012.

20 5.1.2 Scripts Catch and weight at age, natural mortality, and 5 age structured fish abundance indices were used to run ICA. Total catch at age data were compiled by summing catch at age matrices from Bulgaria, Romania, Russia, Turkey and Ukraine. 5 age structured indices were used for deriving the ICA estimates: CPUE from Bulgarian, Crimea and Turkish commercial sprat fleets and relative fish abundance indices from the Romanian Pelagic Trawl Survey(RPTS), and Bulgarian Acoustic survey (BAS) Input data and Parameters Output Generated by ICA Version 1.4 SPRAT Catch in Number AGE x 10 ^ 6 Catch in Number AGE x 10 ^ 6 Predicted Catch in Number AGE x 10 ^ 6

21 Weights at age in the catches (Kg) AGE Weights at age in the catches (Kg) AGE Weights at age in the stock (Kg) AGE Weights at age in the stock (Kg) AGE

22 Natural Mortality (per year) AGE Natural Mortality (per year) AGE Proportion of fish spawning AGE Proportion of fish spawning AGE

23 AGE-STRUCTURED INDICES Bul --- AGE x 10 ^ 3 Bul --- AGE x 10 ^ 3 Crimea AGE ******* ******* x 10 ^ 3 Crimea AGE ******* ******* x 10 ^ 3

24 Rom survey AGE ******* ******* ******* ******* ******* ******* ******* ******* x 10 ^ 3 Turkey AGE ******* ******* ******* ******* ******* ******* x 10 ^ 3 BG acoustic AGE ******* ******* ******* ******* ******* ******* ******* ******* Results ICA was run assuming a constant selection pattern in (Fig ) with reference F at age 2 and Selection at the last real age (S4) equal 1. The results of the ICA show a reasonable fit to observation data (Fig , Fig , Fig ). Fitting to RPTS and BAS data are not shown. The overall fit and partial SSR converged to unique minima (Fig ). Retrospective analyses show some pattern of slightly overestimating F and underestimating recruitment and SSB (Fig ). Analyses of the main population parameters (abundance, catch, fishing mortality, Fig ) indicate that the sprat stock has recovered from the low level observed in the 1990s due to good recruitment in and the biomass and catches have gradually increased over the 1990s and during the 2000s reached levels comparable to the previous periods of high abundance. The stock estimates reveal the cyclic nature the sprat population dynamics. The years with strong recruitment were followed by years of low to medium

25 recruitment which leads to corresponding changes in the Spawning Stock Biomass (SSB). High fishing mortalities (F 1-3) were observed in and and In 2011 the highest ever total catch of t was recorded due mainly to the intensive development of the Turkish sprat fishery. Over years the levels of biomass and catches were comparable with the highest figures reported, but in a decrease in recruitment becomes evident. In catches dropped more than 3 times, and SSB is estimated at the level of about t. After 2013 catch and biomass started rising again reflecting the positive influence of strong year-classes Fig Sprat in GSA 29. Trajectories of the total Sum of Squared Residuals (SSR) and the partial SSRs of the two tuning fleets as functions of the reference F from the ICA final model Age (y) Fig Sprat in GSA 29. Selection pattern estimated by the separable ICA model.

26 Fig Sprat in GSA 29. Time-series of estimated and observed abundance-at-age and age-structured Bulgarian CPUE (best fit is given by linear relationships and r2 are displayed): (a) Age 1. (b) Age 2. (c) Age 3. (d) Age 4. Figure Sprat in GSA 29. Time-series of estimated and observed abundance-at-age and agestructured Crimea CPUE (best fit is given by linear relationships and r2 are displayed): (a) Age 1. (b) Age 2. (c) Age 3. (d) Age 4.

27 Figure Sprat in GSA 29. Time-series of estimated and observed abundance-at-age and agestructured Turkish CPUE (best fit is given by linear relationships and r2 are displayed): (a) Age 1. (b) Age 2. (c) Age 3. (d) Age 4.

28 A B C Figure Sprat in GSA 29. Retrospective anlyses.. SSB and catch are in tonnes, recruitment in 1000s individuals.

29 Fig Sprat in GSA 29. Time-series of sprat population estimates: A) recruitment (line) and SSB (grey); B) landings (grey) and average fishing mortality (ages 1-3 line). SSB and catch are in tonnes, recruitment in 1000s individuals. Diagnostics from the ICA model: Fishing Mortality (per year) AGE Fishing Mortality (per year) AGE

30 Population Abundance (1 January) AGE x 10 ^ 9 Population Abundance (1 January) AGE x 10 ^ 9 Weighting factors for the catches in number AGE

31 Predicted Age-Structured Index Values Bul Predicted AGE x 10 ^ 3 Bul Predicted AGE x 10 ^ 3 Crimea Predicted AGE ******* ******* x 10 ^ 3 Crimea Predicted AGE ******* ******* x 10 ^ 3 Rom survey Predicted AGE ******* ******* ******* ******* ******* ******* ******* ******* x 10 ^ 3

32 Turkey Predicted AGE BG acoustic Predicted AGE ******* ******* ******* ******* ******* ******* ******* ******* Fitted Selection Pattern AGE Fitted Selection Pattern AGE STOCK SUMMARY

33 і Year і Recruits і Total і Spawningі Landings і Yield і Mean F і SoP і і і Age 0 і Biomass і Biomass і і /SSB і Ages і і і і thousands і tonnes і tonnes і tonnes і ratio і 2-3 і (%) і No of years for separable analysis : 7 Age range in the analysis : Year range in the analysis : Number of indices of SSB : 0 Number of age-structured indices : 5 Parameters to estimate : 40 Number of observations : 271 Conventional single selection vector model to be fitted PARAMETER ESTIMATES іparm.і і Maximum і і і і і і Mean of і і No. і і Likelh. і CV і Lower і Upper і -s.e. і +s.e. і Param. і і і і Estimateі (%)і 95% CL і 95% CL і і і Distrib.і Separable model : F by year Separable Model: Selection (S) by age

34 Fixed : Reference Age Fixed : Last true age Separable model: Populations in year Separable model: Populations at age Age-structured index catchabilities Bul Linear model fitted. Slopes at age : 22 1 Q.1878E E E E E E Q.5154E E E E E E Q.6200E E E E E E Q.3573E E E E E E-02 Crimea Linear model fitted. Slopes at age : 26 1 Q.3717E E E E E E Q.8445E E E E E E Q.1738E E E E E E Q.5972E E E E E E-02 Rom survey Linear model fitted. Slopes at age : 30 1 Q.1208E E E E E E Q.2966E E E E E E Q.4903E E E E E E-02 Turkey Linear model fitted. Slopes at age : 33 1 Q.1013E E E E E E Q.3891E E E E E E Q.5589E E E E E E Q.2674E E E E E E-02 BG acoustic Linear model fitted. Slopes at age :

35 37 1 Q.4036E E E E E E Q.2108E E E E E E Q.4417E E E E E E Q.5261E E E E E E-03 RESIDUALS ABOUT THE MODEL FIT Separable Model Residuals Age AGE-STRUCTURED INDEX RESIDUALS Bul --- Age Bul --- Age Crimea Age ******* *******

36 Crimea Age ******* ******* Rom survey Age ******* ******* ******* ******* ******* ******* ******* ******* Turkey Age ******* ******* ******* ******* ******* ******* BG acoustic Age ******* ******* ******* ******* ******* ******* ******* ******* PARAMETERS OF THE DISTRIBUTION OF ln(catches AT AGE) Separable model fitted from 2010 to 2016 Variance Skewness test stat

37 Kurtosis test statistic Partial chi-square Significance in fit Degrees of freedom 14 PARAMETERS OF THE DISTRIBUTION OF THE AGE-STRUCTURED INDICES DISTRIBUTION STATISTICS FOR Bul Linear catchability relationship assumed Age Variance Skewness test stat Kurtosis test statisti Partial chi-square Significance in fit Number of observations Degrees of freedom Weight in the analysis DISTRIBUTION STATISTICS FOR Crimea Linear catchability relationship assumed Age Variance Skewness test stat Kurtosis test statisti Partial chi-square Significance in fit Number of observations Degrees of freedom Weight in the analysis DISTRIBUTION STATISTICS FOR Rom survey Linear catchability relationship assumed Age Variance Skewness test stat Kurtosis test statisti Partial chi-square Significance in fit Number of observations Degrees of freedom Weight in the analysis DISTRIBUTION STATISTICS FOR Turkey Linear catchability relationship assumed Age

38 Variance Skewness test stat Kurtosis test statisti Partial chi-square Significance in fit Number of observations Degrees of freedom Weight in the analysis DISTRIBUTION STATISTICS FOR BG acoustic Linear catchability relationship assumed Age Variance Skewness test stat Kurtosis test statisti Partial chi-square Significance in fit Number of observations Degrees of freedom Weight in the analysis ANALYSIS OF VARIANCE Unweighted Statistics Variance SSQ Data Parameters d.f. Variance Total for model Catches at age Aged Indices Bul Crimea Rom survey Turkey BG acoustic Weighted Statistics Variance SSQ Data Parameters d.f. Variance Total for model Catches at age Aged Indices Bul Crimea Rom survey Turkey BG acoustic

39 5.2 State-Space Assessment Model (SAM) The basic state-space assessment model (SAM) is described in Nielsen & Berg (2014). The method was implemented using the online webpage interface on Model assumptions In SAM, the states (fishing mortalities and abundances-at-age) are constrained by the survival equation and follow a random walk process. The variances of the random-walk processes on abundances and fishing mortalities are parameters estimated by the model. SAM is a fully statistical model in which all data sources (including catches) are treated as observations, assuming a lognormal observation model. The corresponding variances, so-called observation variances, are also parameters estimated by the model. Observations variances can be used to describe how well each data source is fitted in the model and effectively correspond to the internal weight given by the model to the difference data sources. The other parameters estimated are the catchabilities of the surveys. Uncertainties (standard errors) are estimated for all parameters and for all states (Fs and Ns) Scripts Input data, Scripts and assessment models are available on (stock name BSS) Input data and Parameters The assessment model is fitted to catch-at-age data for ages 0 to 4 (plus group) for the period 1997 to 2016 and three tuning indices (fig xx): 1) a nominal CPUE from Bulgaria commercial fleet from 1997 to 2016; 2) a nominal CPUE from Turkish commercial fleet from 2003 to 2010; and 3) a nominal CPUE covering the Crimean region from 1997 to 2016, removing , since the age distribution was incomplete. Two fishery independent survey were available but not included in the SAM assessment model due to really low internal consistency (fig xxx)..

40 Figure : Sprat, GSA 29. Trend in numbers at age for the tuning indices available. Figure : Sprat, GSA 29. Internal consistency for the Romanian survey (left plot) and the Turkish survey (right plot). Compared to ICA assessment model, the plus group was reduced to 4+ instead of 5+ due to 0 catches of age 5 for most of the time series. The maturity and mortality parameters used are shown in table xxx and are the same used in the ICA model:

41 Table : Sprat, GSA 29. Maturity and mortality vector used in the SAM assessment model Maturity Mortality The catch-at-age and weight-at-age matrix are shown in figure xx: Figure : Sprat, GSA 29. Proportion of Catch-at-age and weight-at-age in the catches from 1997 to An autoregressive model was used to model the correlation of fishing mortality across ages Overall settings of the final SAM assessment are: # Configuration saved: Fri Dec 8 11:30: $keylogfsta (Coupling of the fishing mortality states (normally only first row is used) $corflag (Correlation of fishing mortality across ages (0 independent, 1 compound symmetry, or 2 AR(1)) 2 $keylogfpar (Coupling of the survey catchability parameters (normally first row is not used, as that is covered by fishing mortality)) $keyvarf (Coupling of process variance parameters for log(f)-process (normally only first row is used))

42 $keyvarlogn (Coupling of process variance parameters for log(n)-process) $keyvarobs (Coupling of the variance parameters for the observations.) $obscorstruct (Covariance structure for each fleet ("ID" independent, "AR" AR(1), or "US" for unstructured). "ID" "ID" "ID" "ID" $fbarrange (lowest and higest age included in Fbar) 1 3 $fixvartoweight (If weight attribute is supplied for observations this option sets the treatment (0 relative weight, 1 fix variance to weight) Results SSB as resulting from the SAM assessment model is highly fluctuating, from t at the beginning of the time series and about t in The SSB value estimated for 2016 is t (CI: ), with the last 3 years oscillating around an average value of t. Recruitment show a slow but steady increasing trend frfom the beginning of the time seires, with the last year being the highest of the whole time series (R = 364,680,885). The Fbar(1-3) estimates widely oscillate around an average value of (average ) from as low as in 2008 to values as high as in The Fbar value for 2016 is equal to The confidence intervals of all quantities are low on the overall, but they increase in the most recent period (fig xxx). Figure : Sprat, GSA 29. Left: Spawning stock biomass. Middle: Average fishing mortality (Fbar 1-3). Right: Yearly recruitment. Estimates and point wise 95% confidence intervals are shown by line and shaded area for all quantities. Overall results are shown in table xxx; estimated numbers-at-age and fishing mortality-at-age are shown in tables xxx and xxx.

43 Table : Sprat, GSA 29. Summary table for sprat results from SAM model. Year R(age0) Low High SSB Low High Fbar(1-3) Low High TSB Low High E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E

44 Table : Sprat, GSA 29. F-at-age table for sprat results from SAM model. Year Age0 Age1 Age2 Age3 Age

45 Table : Sprat, GSA 29. Numbers-at-age table for sprat results from SAM model. Year Age0 Age1 Age2 Age3 Age Diagnostic plots do not show any particular pattern in the residuals, and the model fit all the data reasonable well (fig xx). Figure : Sprat, GSA 29. SAM fitting of the catch at age data (left) and for the Bulgaria CPUE (right).

46 Figure : Sprat, GSA 29. SAM fitting to the Turkish CPUE (left) and to the Crimean region CPUE (right) Retrospective analysis, comparison between model runs, sensitivity analysis, etc. Retrospective analysis was carried out to test the robustness of the assessment: no retrospective pattern is detected (fig ). Figure : Sprat, GSA 29. Retrospective analysis results for SSB (left), Fbar (middle) and recruitment (right). A leave-one-out sensitivity analysis was carried out as well, to test the effect of the inclusion of the different tuning index in the assessment models. The results show that SSB and Recruitment are quite stable to the inclusion of new indices of abundance. On the other hand, the perception of F change, resulting in higher values whenever the Bulgarian CPUE is included in the assessment model (figure , middle).

47 Figure : Sprat, GSA 29. leave-one-out sensitivity analysis results for SSB (left), Fbar (middle) and recruitment (right). The model results were compared to last year s stock assessment and the assessment carried out during STECF EWG 17-14: the outcomes of the comparison are shown in the plots below Assessment quality The quality of data for Black Sea sprat in 2017 is considered acceptable to perform a reliable assessment. However, two of the surveys had to be removed from the dataset due to poor internal consistency between age classes. Besides, the use of nominal CPUE as tuning index for small pelagic is not advisable and might lead to a wrong perception of the stock status. The SAM results were compared with the STECF 2017 results from ICA and with last year assessment (ICA as well) (figures xxx). We noticed that the SSB reported in both the GFCM and the STECF report is not at time of spawning, which happens at mid-year: it accounts in fact for half of the natural mortality setting the value of M at age 0 = 1.28/2, but the proportion of F before spawning was set = 0 in the input file. For the comparison therefore we had to manually calculate the mid-year spawning stock biomass.

48 Figure : Sprat, GSA 29. Model comparison for the main stock variables (top: Fbar 1-3; middle: Mid Year SSB; bottom: recruitment). Exploitation rate between the ICA carried out during STECF 2017 and the SAM assessment was compared as well. Confidence intervals are available from the SAM model only. As shown in the plot below, the ICA results lay on the lower bound of the SAM estimates and are -most of the time- within the SAM confidence intervals. The overall perspective, however, is quite different: according to ICA, from 1997 the stock has been above the 0.4 reference point only 15% of the time, while according to the SAM model this percentage increase to 70%. Figure : Sprat, GSA 29. Exploitation rate from SAM model and ICA The horizontal blue line show the Patterson s reference point of 0.4.