Falkland Island Fisheries Department. Castelo (ZDLT1) Falkland Islands. Dates 9/02/ /02/ May

Transcription

1 Falkland Island Fisheries Department Vessel Flag Castelo (ZDLT1) Falkland Islands Dates 9/02/ /02/2009 Author Ignacio Payá Scientific Crew Ignacio Payá, Sarah Hearne, Paul Brickle and Joost Pompert - May

2 SUMMARY A research survey was conducted in the Loligo box of the Falkland Islands shelf on board F/V Castelo between the 9 th and 23 rd of February hauls were made in selected localities with a total Loligo catch of 187 tonnes. A DST-CTD logger was attached to the net to collect oceanographic data. The standardised biomass was estimated at tonnes, which was 1.7, 5.7, 1.1, 0.4 times the biomass estimated in 2008, 2007, 2006 and 2005 February surveys, respectively. The biomass was composed mainly (61%) by females, as happened in February The average mantle length was 11.6 cm, which was larger than in the three previous February surveys. Following the general spatial pattern of previous first seasons, Loligo were mainly concentrated in the southern area, however inside this area Loligo were more concentrated in south-east part and at deeper waters than in previous February surveys. Loligo spatial distributions within the southern area have changed by season: highest concentrations were located in the north part in 2005 and 2006; in the middle part in 2007 and 2008; and in the south-east in During the survey 2009, there was a significant relationship between Loligo presence and bottom salinity: the higher salinity - more presence. There was also a significant relationship between Loligo density and bottom salinity and bottom temperature; higher salinity - higher density and lower temperature - higher density. Bottom salinity and temperature were successfully incorporated in spatial models as covariates of Loligo presence and density, and they were used to improve the estimations of biomass and spatial distributions. The spatial models with oceanographic covariates were statistically better than previous models without oceanographic covariates. 1

3 INDEX I. INTRODUCTION...3 II. OBJECTIVES...3 III. METHODOLOGY Sampling design and Biomass estimations Oceanographic sampling and data analyses...7 IV. RESULTS AND DISCUSSION Distribution and catch rates Biological characteristics Biomass estimations without oceanographic covariates Oceanography Row data spatial distribution Spatial distribution by geo-statistical analyses GAM and GLM of Loligo and oceanographic variables a. Loligo presence b. Loligo density Biomass estimations with oceanographic covariates V. CONCLUSIONS VI. ACKNOWLEDGMENT VII. REFERENCES

4 I. INTRODUCTION The current survey is the tenth made since May 2004, when the first scientific survey onboard a commercial trawler was conducted (Roa-Ureta 2004, 2005a, 2005b; Payá and Roa-Ureta 2006 and Payá 2006a, 2007a and 2007b, 2008a and 2008b). The first three surveys were made long before the next fishing season and therefore the biomass at the beginning of the next fishing season had to be estimated using projection models. To avoid natural uncertainty of these projections and any possibility that resource had not fully recruited to the fishing area, since February 2006 the surveys were made just before the beginning of the fishing season. In February 2008, because different vessel had been participated in the survey a vessel standardisation procedure was introduced, which was based on the trawling speed by vessel (Payá 2008a). In July 2008, for first time oceanographic information was collected by data storage tags attached to the trawling net. This showed that Loligo were concentrated in waters to the upper part of the Transient Zone and had not completely entered the Loligo box because it was restricted by the extension of these waters (Payá 2008b). II. OBJECTIVES 1. To estimate Loligo biomass available to the bottom trawl fishery along the Loligo box, just before the start the fishing season. 2. To sample the oceanographic conditions (temperature, salinity and density) of the seawater at the bottom. 3. To analyse the relationship between Loligo presence and density and oceanographic variables. III. METHODOLOGY 1. Sampling design and Biomass estimations All fishing activities were performed on the F/V Castelo, a Stanley registered stern trawler with total length of 67.8 m, a beam of 11 m and draught of 5.07 m. The gross registered tonnage is 1321 mt with a net registered tonnage 474 mt. A total of 59 hauls 3

5 were made in locations selected by scientists yielding a total catch of 187 tonnes. Between four and five trawls were conducted each day when the weather allowed. The trawl was a standard Spanish made bottom trawl with a small mesh liner in the codend. The door opening varied from 85 to 115 m with a mean of 103 m, and the horizontal trawl opening ranged from 35 to 47 m with a mean of 43 m depending on the course and trawl speed. Trawl duration varied but was normally between two to three hours. Every fifteen minutes during each tow the bridge officers noted the position, trawl speed, door opening and quantity and quality of the marks observed on the echosounder. The net was hauled on board and lifted into place to allow the catch to flow into one of two fish bins at stern of the trawl deck. The fish bins fed a conveyor system in the factory. A random sample of 150 squid was taken from every trawl and as soon as they were separated by sex and maturity they were measured for length frequency analysis. Additionally, all by-catch species were collected from each trawl by crew members working at the conveyor belt. After the contents of the trawl had been processed, the bycatch was weighed and some species like Illex argentinus, rockcod, icefish and skates were examined in greater detail. The survey covered the whole shelf area of the Loligo box (depths ranging between 80 and 320 m). The survey consisted of 14 transects, with several trawls on each transect depending on the width of the shelf in the area (Fig. 1). The biomass was estimated using geo-statistical methods. The biomass was the product of the probability of presence, the mean density and the total area (Roa-Ureta 2004, 2005). The probability of presence was estimated by Binomial logit-normal models, using georglm R package (Christensen and Ribeiro Jr, 2005). The mean density was estimated using the likelihood-based methods of Diggle et al. (1998 and 2003) and the geor R package (Ribeiro Jr and Diggle, 2001). 4

6 50 S T14 T13 51 S T12 T11 Latitude 52 S T7 T8 T9 T10 T6 53 S T1 T2 T3 T4 T5 54 S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Longitude Fig. 1. Adaptive sampling design of 59 stations (red dotted lines) throughout 14 transects (green lines) of the Loligo survey in February White polygons represent the hard bottoms. 5

7 Since the initial formulation of the procedures to estimate the biomass (Roa-Ureta 2004, 2005), several improvements have been introduced: 1) Sampling design with a fixed and an adaptive sample component (Payá 2006b). The fixed sample design was introduced to cover all Loligo box, and the adaptive sampling to increase the precision of estimations in high-density localities (hot pots). In the initial surveys the adaptive sampling design was applied only and therefore the sampling in low densities localities was scarce. 2) Anisotropic analyses were introduced in the geostatistic analysis of density and presence (Payá, I. 2006b). The anisotropy conditions happen when spatial correlation changes by spatial orientation. For example, the anisotropic condition could occur in the southern area because Loligo is distributed along depth contours that have a clear East-West orientation. 3) The biomass estimations were standardised using the trawling speed by vessel (Payá 2008a). The standardisation was introduced because different vessels have been used in the surveys. 4) The area occupied by the stock was estimated as the product of the surveyed area and the proportion of 5*5 km grid squares with Loligo presence (introduced in this report). The surveyed area was estimated by drawing a polygon that enclosed the stations locations. The surveyed area was estimated for every season, because in some seasons Loligo were more extended to the coastal waters than in others. In previous biomass estimations the surveyed area was equal to the fishing grounds area, which was assumed to be constant between seasons. 5) Improve of biomass and spatial distribution estimations using oceanographic variables as covariates (present report). In previous biomass estimations it was assumed that both density and presence had not trend and depend on spatial coordinates only. In the case of the spatial model for density, the density trend was modelled with salinity and temperature as covariates and in the case of the spatial model of presence, the presence trend was modelled with salinity only. After the spatial models were fitted, the Loligo spatial predictions were based on the spatial coordinates and the oceanographic covariates. As oceanographic data were collected in the same Loligo stations, the oceanographic covariates used to predict the Loligo density and presence in the grid (kriging) were the 6

8 estimations of the oceanographic variables in the grid (kriging) previously done by geo-statistical analyses of the oceanographic variables (see next section on this report). In order to have biomass estimations to be used in depletion models during the inseason stock assessment, biomasses were also calculated for the southern area and the centre-north area. To estimate the mean density and presence by area, the spatial predictions, done using the spatial model fitted to the data of the whole area, were divided by area and then the averages were calculated. 2. Oceanographic sampling and data analyses. In order to collect oceanographic information a Data Storage Tag (DST) CTD (Star- Oddi, was attached to the upper rope of the net mouth. DST CTD is a compact microprocessor-controlled conductivity, temperature and depth recorder with electronics housed in a waterproof housing. In every haul the recorders were removed from the net and put in the communication box to retrieve the data and to program the recorders for the next haul. DST recorded CTD every 2 minutes during each trawl. To compare with Loligo density data, which were estimated every 15-minute intervals, the arithmetic mean of CTD data were calculated by 15-minute intervals. To compare with Loligo presence data, which were computed by 5x5 Km cells, the arithmetic mean of CTD data were calculated by 5x5 Km cells. Sea water densities were calculated using Excel Macro DENSATP(S,T,P), which calculates density at a given salinity, temperature and hydrostatic pressure. This macro used the equation of state from Millero and Poisson (1981). The pressures were calculated using macro PRESS(D,LAT), which estimates pressure at a given depth and latitude. This macro is based on Saunders (1981). Both macros are available at the webpage of Dr. Edward T. Peltzer ( 7

9 With good precision but limited accuracy the DST-CTD had to be post-calibrated using a proper CTD. The DST-CTD and the proper CTD were simultaneously deployed in different localities that covered a wide range of salinities and temperatures: Murrell River, Moody Brook, West Stanley Harbour (Beaver Hangar), FIPASS, East Stanley Harbour (Boxer Bridge) and Surf Bay coast. The oceanographic information was analyzed using the likelihood-based geostatistic methods of Diggle et al. (1998 and 2003) and the geor R package (Ribeiro Jr and Diggle, 2001). The same spatial resolution (5x5 Km) used in the Loligo density geostatistic analysis was used for the oceanographic variables. The relationship between Loligo presence and environmental variables was analysed using GAM with binomial error distribution and logit link function. The models were compared using the Akaike Information Criterion (AIC). Based on the best GAM a GLM was fitted using a second order polynomial for the environmental variable. The same procedure was used to analyse the relationship between Loligo density and environmental variables, but gamma error distribution and log link function were used in GAM and GLM. 8

10 IV. RESULTS AND DISCUSSION 1. Distribution and catch rates Loligo were concentrated in the south area of the Loligo box, and were scarce in the middle and north areas (Fig. 2). 50 S 51 S Latitude 52 S 53 S CPUE (tonnes/h) 0 to to to to to S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Longitude Fig.2. Loligo CPUE (tonnes/hour) observed during February 2009 survey. 9

11 Most of the trawls done in the middle and the northern area had zero or very small Loligo catches (Fig. 3). In the southern area, Loligo was concentrated to the east and south-east of Beauchene Island (Fig. 4) S 50.7 S 50.9 S 51.1 S 51.3 S Latitude 51.5 S 51.7 S 51.9 S 52.1 S 52.3 S CPUE (tonnes/h) 0 to to to to to S W W W W W Longitude Fig. 3. CPUE (tonnes/h) of L. gahi observed during the February 2009 in the Centre- North area. 10

12 52.6 S 52.8 S Latitude 53 S 53.2 S CPUE (tonnes/h) 0 to to to to to W 60 W 59.5 W 59 W 58.5 W Longitude Fig. 4. Loligo CPUE (tonnes/h) observed during February 2009 Survey in the southern area of Loligo box (blue line). Rockcod (Patagonotothen ramsayi) was abundant in the northern and central areas of the Loligo box, with a trend to increase toward northern and deeper waters (Fig. 5). As in previous February surveys, there was an inverse relationship between the proportions of Loligo and rockcod in total catches (Fig. 6). Therefore, the possible species interference in the acoustic mark identification was small. 11

13 50 S 51 S Latitude 52 S 53 S Proportion of RockCod in Catch 0 to to to to to 1 54 S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Longitude Fig. 5. Proportion of Rockcod in total catch at the initial positions of each trawl. 12

14 1 Rockcod Proportion Loligo Proportion Fig. 6. trawl. Relation between proportion of Loligo and rockcod in total catch by each 2. Biological characteristics The average Loligo mantle length was 11.6 cm for both sexes combined and 11.4 cm for females and 12.2 cm for males (Fig. 7). Squid sizes in 2009 were similar to the sizes in 2008, but greater than the sizes in previous February surveys. The largest Loligo were found mainly in the deepest waters in the southern area (Fig. 8). Females were more abundant than males (female proportion >0.6) along the Loligo box, and only in some trawls in the southern area the female proportions were low (Fig. 9). Female proportions in the localities of the highest concentrations varied from to

15 FEMALE 0.2 FREQUENCY MANTLE LENGTH (cm) MALE 0.2 FREQUENCY MANTLE LENGTH (cm) COMBINED 0.2 FREQUENCY MANTLE LENGTH (cm) Fig. 7. Loligo mantle length frequency by sex and sexes combined found in February surveys. 14

16 50 S 51 S Latitude 52 S 53 S Average Mantle Length (cm) 7 to to to to to S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Longitude Fig. 8. Loligo average mantle length at the start positions of each trawl. 15

17 50 S 51 S Latitude 52 S 53 S Female Proportion 0 to to to to to 1 54 S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Longitude Fig. 9. Loligo female proportion found at the start positions of each trawl. 16

18 3. Biomass estimations without oceanographic covariates The standardised biomass available during the survey February 2009 was estimated at tonnes, with a coefficient of variation of 38% (Tables 1 and 2). The raw biomass was estimated at tonnes and the correction factor for the F/V Castelo at The biomass estimations by sex were more precise in males than in females (CV in Table 2). The 2008 biomass was composed by 61% of females as happened in The whole 2009 standardised biomass corresponded to 570 millions of individuals, which had a mean body weight similar than squids in 2008 but greater than in the others February surveys. Table 1. Main results of February surveys by year Biomass (tonnes) Correction factor Standardised Biomass (tonnes) Female Proportion Mean Density (g/cm^2) Standardised Number (million) Area occupied by the stock Fishing Grounds Area (km^2) Mean Body Mass (g) Table 2. Main results of 2009 February survey by sex. Biomass by sex. The total is not the simple sum but the result of geostatistic analysis done with both sexes combined. Total Female Male Area occupied by the stock (km^2) Biomass (tonnes) Standard Error Biomass (tonnes) CV Biomass % Mean Body Mass (g) Number (million) SD(Number) (million) Var(Number) (million^2)

19 The biomasses of previous surveys recalculated using different areas (polygons) by year were greater than the biomasses estimated using a fixed area of fishing grounds (Table 3). The differences in the areas by year were because of the changes in Loligo distribution and the adaptable sampling design of the surveys. Table 3. Standardised biomasses estimated with different areas Area of Polygons (km^2) Area of Fishing Grounds Area (km^2) Biomass with area of polygones (tonnes) Biomass with fishing grounds area (tonnes) Biomass ratio The spatial statistics analysis for the presence/absence data, first component of biomass estimation, showed that during February 2009 Loligo were present in 76% of surveyed area (Table 4). This is the highest figure estimated in all the surveys. Anisotropy condition was found in the data. The anisotropy angle was estimated at 0 degrees and the ratio at 1.22, which are lower than anisotropic parameters of previous seasons. Table 4.- Descriptive statistics and parameters of the spatial Loligo presence/absence process in February surveys Number of Localities (5 x 5 km squares) Area of locality Total Number of Trials Total Number of Successes Spatial AC Function Gaussian Whitle-Matern Whitle-Matern Whitle-Matern Whitle-Matern Family Binomial Binomial Binomial Binomial Binomial Link Function Logit Logit Logit Logit Logit Number of parameters Non Spatial Model Log-likelihood Non Spatial Model AIC Non Spatial Model Number of parameters Spatial Model Log-likelihood Spatial Model AIC Spatial Model Kappa (fixed) Inf Tau^sq (nugget) (fixed) Sigma^sq (sill) Phi (range) (km) Beta Spatial model Isotropy Isotropy Anisotropy Anisotropy Anisotropy anisotropy angle (degrees) anisotropy ratio Kriging Mean p Mean Interpolation SD of p CV Mean Interpolation p

20 The spatial autocorrelation was fitted to the Whitle-Matern function, 2005 was the only season when Gaussian model was fitted to the data because the Whitle-Matern function did not converge (Table 4). Loligo was less concentrated in 2009 than in 2005 but more concentrated than in the others years (Fig. 10). Correlation Year Kilometers Fig. 10. Spatial correlation for presence/absence information for February surveys. The spatial correlation for the presence proportion ended at 50-kilometre distance in 2009 and 2008, at 150-kilometre distance in 2007 and 2006 and at 20-kilometre distance in Following the general spatial pattern of previous seasons, Loligo was present mainly in the southern area during 2009 (Fig. 11). However, within the southern area Loligo presence was more concentrated in south-east part, at deeper waters than in previous February surveys. The distribution of Loligo presence in the southern area have showed important changes by year that are likely related with the intensity and timing of Loligo immigration to this area. In 2005 and 2006 the concentration areas were closer to the north of the southern area, where it is supposed that Loligo enter to the area. In 2007 the presence area was located mostly in the center and in the southern part of the southern area, that year the biomass was very low and the immigration was delayed by three weeks. In 2008 Loligo were present in the south-east part, but more groups arrived later during the fishing season. 19

21 Falkland Islands Fisheries Department Presence 4250 Northing (km) 4300 Presence Northing (km) Presence Presence Northing (km) Northing (km) 550 Easting (km) Easting (km) Easting (km) Easting (km) 4250 Presence Northing (km) Easting (km) Fig. 11. Loligo presence proportion estimations for February survey from 2006 to

22 The spatial statistics analysis of the density, second component of biomass estimation, showed that February 2009 survey had similar number of positive observations than in 2006 and greater than in the others seasons (Table 5). In 2009, the female mean density (1.87 g/m 2 ) was greater than in males. Both sexes had anisotropic conditions, which anisotropic parameters (angle and ratio) lower than in previous seasons. The correlation was fitted to Gaussian models, as in 2005 and The spatial correlations of density showed that Loligo were more concentrated than in 2008 but less than in the others seasons (Fig. 12). Correlation Year Kilometers Fig. 12. Spatial correlation for positive density information for February surveys. The spatial distributions of density by sex were similar and showed that Loligo densities were very low along the central and northern areas of the Loligo box and that the highest densities were located in south-east part of the southern area close to Beauchene Island (Figs 13 and 14). 21

23 Table 5. Descriptive statistics and parameters of the spatial Loligo density process in February surveys Female Male Female Male Female Male Female Male Female Male Number of Observations Spatial AC Function Gaussian Gaussian Whitle-Matern Whitle-Matern Whitle-Matern Whitle-Matern Gaussian Gaussian Gaussian Gaussian AIC Non Spatial Model AIC Spatial Model Lambda Kappa Inf Inf inf inf inf inf Tau^sq (nugget) (g^2/m^4) Sigma^sq (sill) (g^2/m^4) Phi (range) (km) Beta (g/cm^2) Spatial model Isotropy Isotropy Isotropy Isotropy Anisotropy Anisotropy Anisotropy Anisotropy Anisotropy Anisotropy Anisotropy angle (degrees) Anisotropy ratio Kriging Mean (g/m^2) Mean Interpolation SD (g/m^2) Kriging Beta (g/m^2) SD BT-Kriging Beta (g/m^2) CV BT-Kriging Beta

24 Northing (km) Density (g/m^2) Easting (km) Fig. 13. Female density estimations for February 2009 survey. Northing (km) Density (g/m^2) Easting (km) Fig. 14. Male density estimations for February 2009 survey. Please note different scale than in Fig. 13, this is in order to be able to compare the spatial distribution rather than the density levels. The spatial distribution of density for both sexes combined showed the same general pattern by year, that is highest densities are located in the southern area (Fig. 15). 20

25 There were big differences in the magnitude of densities by year; this fact forced using different scales in the plots. The lowest densities were found in 2007 and the highest in The density values during 2009 were similar to The spatial distributions in the southern area changed by season: the highest concentrations were located in the north part during 2005 and 2006; in the middle in 2007 and 2008; and in the south-east in

26 Falkland Islands Fisheries Department Density (g/m^2) 4250 Northing (km) 4300 Density (g/m^2) 4250 Northing (km) Density (g/m^2) 4250 Density (g/m^2) Northing (km) Northing (km) 500 Easting (km) Easting (km) Easting (km) Easting (km) 4300 Density (g/m^2) 4250 Northing (km) Easting (km) Fig. 15. Loligo density (tonnes/km2) in February surveys from 2005 to Please note the different scales used in each plot in order to be able to compare the spatial distribution rather than the density levels. 22

27 4. Oceanography 4.1 Row data spatial distribution The post-calibration was successful. For both salinity and temperature linear relationships between data recorded by DST-CTD and by proper CTD were found (Fig. 16). These linear models were used to adjust the data recorded by the DST-CTD. Salinity y = x R 2 = CTD DST CTD CTD Temperature y = x R 2 = DST CTD Fig. 16. Linear relationships fitted to the data recorded by DST CTD and by the proper CTD during post-calibration. The waters were cooler in deeper and northern stations (Fig. 17) and they showed a clear decreasing trend from shallow to deeper stations in the southern area (Fig. 18). Loligo were more abundant in water with temperatures less than 6.64ºC. 23

28 50 S 51 S Latitude 52 S 53 S Bottom Temperature (C) 4.99 to to to to to S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Longitude Fig. 17. Bottom temperature during February

29 52.6 S 52.8 S Latitude 53 S 53.2 S Bottom Temperature (C) 4.99 to to to to to W 60 W 59.5 W 59 W 58.5 W Longitude Fig. 18. Bottom temperature in the southern area during February The water salinity was higher in deeper and southern stations (Fig. 19) and they showed a clear increasing trend from shallow to deeper stations in the southern area (Fig. 20). Loligo were more abundant in water with salinity greater than psu. 25

30 50 S 51 S Latitude 52 S Salinity (psu) to S to to to to S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Fig. 19. Bottom salinity during February Longitude 26

31 52.6 S 52.8 S Latitude 53 S Salinity (psu) to to to S to to W 60 W 59.5 W 59 W 58.5 W Longitude Fig. 20. Bottom salinity in the southern area during February The waters were denser (Sigma-t) in deeper and southern stations (Fig. 21), with a clear increasing trend from shallow to deeper stations in the southern area (Fig. 22). Loligo were more abundant in water with densities greater than 27.3 Sigma-t. 27

32 50 S 51 S Latitude 52 S 53 S Sigma-t 25.9 to to to to to S 63 W 62 W 61 W 60 W 59 W 58 W 57 W 56 W Longitude Fig. 21. Bottom seawater density (sigma-t) during February

33 52.6 S 52.8 S Latitude 53 S 53.2 S Sigma-t 25.9 to to to to to W 60 W 59.5 W 59 W 58.5 W Longitude Fig. 22. Bottom seawater density (sigma-t) in the southern area during February Spatial distribution by geo-statistical analyses. The plots, produced by geo-statistical analyses, showed that both presence and density of Loligo were similar to the spatial distribution of salinity (Fig. 23). The spatial distribution of water density was also similar to Loligo distribution in the southern area, but no in the central area. The temperatures showed an important gradient from shallow to deeper waters but were less related with Loligo distribution along the longitude. 29

34 Density (g/m^2) 4250 Northing (km) 4250 Presence Northing (km) Falkland Islands Fisheries Department Salinity Northing (km) Temperature (C) 4250 Northing (km) 500 Easting (km) Easting (km) Easting (km) 4250 Sigma-t Northing (km) Easting (km) Easting (km) Fig. 23. Loligo presence and density distribution (top row); bottom temperature and bottom salinity distribution (middle row); and bottom seawater density (down plot), during February

35 4.3 GAM and GLM of Loligo and oceanographic variables. a. Loligo presence The first analysis done included the Loligo presence and the oceanographic and geographic variables. The pair relationships between all variables are shown in Fig BTemp BSalinity BSigma.t Northing Easting Depth LolPresence Fig. 24. Pair plots between Loligo presence (LolPresence), oceanographic variables (BTemp: Bottom temperature; BSalinity: Bottom Salinity; BSigma.t: Bottom seawater density), depth and spatial coordinates (Northing and Easting). 31

36 The transformations (spline) of oceanographic variables in GAM with only one variable are shown in Fig. 25. s(btemp,7.18) BTemp s(bsalinity,2.5) BSalinity s(bsigma.t,2.82) BSigma.t Fig. 25. GAM of Loligo presence and bottom temperature (top plot), bottom salinity (middle plot) and bottom seawater density (down plot). For each variable the GAM were fitted separately. The plots show the smooth transformations (spline) of the variables with their 95% confident intervals (red area). 32

37 The transformations (spline) of a GAM with temperature plus salinity are shown in Fig BTemp s(bsalinity,2.5) s(btemp,1) BSalinity Fig. 26. GAM of Loligo presence with bottom salinity (top plot) plus bottom temperature (down plot). The best GAM was the GAM with salinity only (Tables 6 and 7). The GLM with second degree polynomial of salinity was highly statistically significant (Table 8). Table 6. Comparison of Loligo presence and oceanographic variables based on Akaike Information Criterion (AIC). The model with the lowest AIC is the best. Model Deviance explained AIC Temperature 13.80% 297 Seawater density 8.61% 300 Salinity 14.50% 286 Salinity + Temperature 14.70%

38 Table 7. Main results of GAM of Loligo presence and bottom salinity (BSalinity). Family: binomial Link function: logit Formula: LolBino ~ s(bsalinity) Parametric coefficients: Estimate Std. Error z value Pr(> z ) (Intercept) <2e-16 *** --- Signif. codes: 0 *** ** 0.01 * Approximate significance of smooth terms: edf Ref.df Chi.sq p-value s(bsalinity) e-06 *** --- Signif. codes: 0 *** ** 0.01 * R-sq.(adj) = Deviance explained = 14.5% UBRE score = Scale est. = 1 n = 152 Table 8. Main results of GLM of Loligo presence with bottom salinity (BSalinity). Call: glm(formula = LolBino ~ poly(bsalinity, 2), family = binomial(link = "logit"), data = OceanoLol, na.action = na.omit) Deviance Residuals: Min 1Q Median 3Q Max Coefficients: Estimate Std. Error z value Pr(> z ) (Intercept) < 2e-16 *** poly(bsalinity, 2) ** poly(bsalinity, 2) *** --- Signif. codes: 0 *** ** 0.01 * (Dispersion parameter for binomial family taken to be 1) Null deviance: on 151 degrees of freedom Residual deviance: on 149 degrees of freedom AIC: Number of Fisher Scoring iterations: 4 This is the first time that Loligo presence is related with oceanographic variables, previous studies (Arkhipkin et al and 2005) were based on density (commercial 34

39 CPUE) and not in presence/absence data. Therefore, it is new that Loligo presence is more correlated with salinity than with temperature or seawater density. b. Loligo density. The second analysis done included the Loligo density and the oceanographic and geographic variables. The pair relationships between all these variables are shown in Fig BTemp BSalinity BSigma.t Northing Easting Depth LolDensity Fig. 27. Pair plots between Loligo density (LolDensity), oceanographic variables (BTemp: Bottom temperature; BSalinity: Bottom Salinity; BSigma.t: Bottom seawater density), depth and spatial coordinates (Northing and Easting). 35

40 s(btemp,8.49) BTemp s(bsalinity,8.44) BSalinity s(bsigma.t,8.33) BSigma.t Fig. 28. GAM of Loligo density and bottom temperature (top plot), bottom salinity (middle plot) and bottom seawater density (down plot). For each variable the GAM were fitted separately. The plots show the smooth transformations (spline) of the variables with their 95% confident intervals (red area). 36

41 The transformations (spline) of a GAM with temperature plus salinity are shown in Fig. 28. s(bsalinity,8.35) BSalinity s(btemp,1) BTemp Fig. 29. GAM of Loligo density with bottom salinity (top plot) plus bottom temperature (down plot). The best GAM was the GAM with salinity plus temperature (Tables 9 and 10). The GLM with second degree polynomials of salinity and temperature showed that the second degree of polynomial of temperature was not significant, and therefore the GLM was fitted with bottom temperature and a second degree of polynomial of salinity. This GLM was highly statistically significant (Table 11). Table 9. Statistical comparison of Loligo log density and oceanographic variables based on Akaike Information Criterion (AIC). The model with the lowest AIC is the best. Model Deviance explained AIC Temperature 25% 1535 Seawater density 37% 1469 Salinity 62% 1283 Salinity + Temperature 67%

42 Table 10. Main results of GAM of Loligo density with bottom salinity (BSalinity) plus bottom temperature (BTemp). Family: Gamma Link function: log Formula: LolDensity ~ s(bsalinity) + s(btemp) Parametric coefficients: Estimate Std. Error t value Pr(> t ) (Intercept) <2e-16 *** --- Signif. codes: 0 *** ** 0.01 * Approximate significance of smooth terms: edf Ref.df F p-value s(bsalinity) <2.00E-16 *** s(btemp) E-05 *** --- Signif. codes: 0 *** ** 0.01 * R-sq.(adj) = Deviance explained = 51.8% GCV score = Scale est. = n = 376 Table 11. Main results of GLM of Loligo density with bottom temperature (BTemp) plus bottom salinity (BSalinity). Call: glm(formula = LolDensity ~ BTemp + poly(bsalinity, 2), family = Gamma(link = "log"), data = OceanoLol, subset = LolDensity > 0, na.action = na.omit) Deviance Residuals: Min 1Q Median 3Q Max Coefficients: Estimate Std. Error t value Pr(> t ) (Intercept) *** BTemp E-07 *** poly(bsalinity,2) < 2e-16 *** poly(bsalinity,2) E-07 *** --- Signif. codes: 0 *** ** 0.01 * (Dispersion parameter for Gamma family taken to be ) Null deviance: on 375 degrees of freedom Residual deviance: on 372 degrees of freedom AIC: Number of Fisher Scoring iterations: 9 38

43 Although the model of Loligo density with salinity plus temperature was statistically the best, the contribution of temperature was small. When temperature was added to the model the deviance explained by the model increased in 5% only. Therefore, as the Loligo presence/absence, the Loligo density was also mainly related with salinity. Arkhipkin et al. (2005) postulated that Loligo distribution is limited by the upper part of the Transient Zone waters, which is indicated by salinity of psu. However during the survey Loligo were distributed beyond this limit. This could be explained by the fact that Arkhipkin et al. (2005) did their analyses using commercial CPUE as density variable. The use of commercial data clearly biased the data to the places of high densities. This is not the case in the Loligo survey, where the whole Loligo box was sampled. The highest salinities found in the southern area were greater than the historical salinities recorded by proper CTD in the Falkland Islands waters. Although the linear calibration model was good, it did not include salinities higher than 33.9 psu, and so higher salinities were outside the range of the regression. Therefore, it is recommended to improve the post-calibration model with additional points of higher salinities. 5. Biomass estimations with oceanographic covariates. The standardised biomass in the whole area was estimated at tonnes, which is 92% of the whole biomass estimated without oceanographic covariates (Table 12). However these biomass estimations are not statistically different because both have high CV. The biomasses in the southern and centre-north area were estimated at and 2348 tonnes. The whole biomass was estimated independently by geostatistic method, and so it is not the simple sum of biomasses by area. The only way that the sum of biomass by area could be equal to the whole biomass, is to assume that mean density and mean proportion of presence is the same in both area, and then calculate the biomass by area using the ratio of areas. But this is incorrect because the mean density and mean presence are different by area. 39

44 Table 12. Biomasses estimated with oceanographic trend by area and whole (STD: Standard Deviation; CV: Coefficient of Variation). The whole biomass was estimated independently by geo-statistic method, and so it is not the simple sum of biomasses by area. Centre-North South Whole Mean Density (tonnes/km 2 ) STD Density CV Density Area (Km 2 ) Mean Presence Proportion STD Presence Proportion CV Presence Proportion Area occupied by stock Biomass (tonnes) Fishing Power Correction factor Standardised Biomass (tonnes) STD Standardised Biomass CV Standardised Biomass The models with oceanographic trend were statistically better (lower AIC) than the models without oceanographic covariates (Table 13). Table 13. Comparison between spatial models with and without oceanographic covariates; based on Akaike Information Criterion (AIC). The model with the lowest AIC is the best. Without Ocenographic covariates AIC With Ocenographic covariates Variable Presence Density The spatial distributions of the presence and density were better than the spatial distributions done with models without oceanographic covariates; they showed in a better way the place of high density concentration where later the fishing fleet caught Loligo (Fig. 30). 40

45 4300 Presence 4250 Northing (km) 4300 Presence 4250 Northing (km) Falkland Islands Fisheries Department Density (g/m^2) 4250 Northing (km) Density (g/m^2) 4250 Northing (km) 500 Easting (km) Easting (km) Easting (km) Fig Easting (km) Loligo presence (top plots) and density (down plots) estimated by spatial models with (right plots) and without (left plots) oceanographic covariates. 41

46 V. CONCLUSIONS 1. Loligo biomass during the F/V Castelo survey, between 8 th and 23 rd of February of 2009, was estimated at tonnes. This biomass was 1.7, 5.7, 1.1, 0.4 times the biomass estimated in 2008, 2007, 2006 and 2005 February surveys, respectively. 2. The biomass was composed mainly (61%) by females, as happened in February The average mantle length was 11.6 cm, which was larger than in the three previous February surveys. 4. Following the general spatial pattern of previous first seasons, Loligo were mainly concentrated in the southern area. However, within the southern area Loligo were more concentrated in south-east part and at deeper waters than in previous February surveys. 5. Loligo spatial distributions within the southern area have changed by season: highest concentrations were located in the north part in 2005 and 2006; in the middle part in 2007 and 2008; and in the south-east in There was significant relationship between the Loligo presence and the bottom salinity: Higher salinity more Loligo presence. 7. There was a significant relationship between the Loligo density and bottom salinity and bottom temperature: Higher salinity higher Loligo density and lower temperature higher Loligo density. 8. Bottom salinity and temperature were successfully incorporated in spatial models as covariates of Loligo presence and density, and they were used to improve the biomass estimations and spatial distributions. 42

47 9. The spatial models with oceanographic covariates were statistically better than previous models without oceanographic covariates. VI. ACKNOWLEDGMENT This study could not be done without the collaboration of the captain and crew of the F/V Castelo. The captain and bridge officers actively participated in the identification and classification of Loligo acoustic marks. The crew helped attaching and removing the data storage tags from the net in every haul and facilitated the biological sampling in the factory. The survey was conducted in a very friendly environment. Dr. Paul Brickle and Mr. Joost Pompert were onboard during the first 4 days of the survey. Dr. Vlad Laptikhovsky participated in the post-calibration of DST-CTD with a proper CTD. VII. REFERENCES Arkhipkin, A.I., R. Grzebielec, A.M. Sirota, A.V. Remeslo, I.A. Polishchuck & D.A.J. Middleton The influence of seasonal environmental changes on ontogenetic migrations of the squid Loligo gahi on the Falkland shelf. Fisheries Oceanography, 13:1-9. Arkhipkin, A.I., D.A.J. Middleton. R., A.M. Sirota. & R. Grzebielec The effect of Falkland Current inflows on offshore ontogenic migrations of the squid Loligo gahi on the southern shelf of the Falkland Islands. Estuarine, Coastal and Shelf Science 60: Diggle PJ, Tawn JA, Moyeed RA Model based geostatistics (with discussion). Applied Statistics. 47: Diggle PJ, Ribeiro PJ, Christensen OF An introduction to model-based geostatistics. In J. Moller (Ed.), Spatial Statistics and Computational Methods, vol. 173, Lecture Notes in Statistics, Springer.Diggle PJ, Ribeiro PJ, Christensen OF An introduction to model-based geostatistics. In J. Moller (Ed.), Spatial Statistics and Computational Methods, vol. 173, Lecture Notes in Statistics, Springer. Millero, F.J. and A. Poisson International one-atmosphere equation of state of seawater. Deep Sea Research Part A. Oceanographic Research Papers, Volume 28, Issue 6, June 1981, Pages Payá, I and R., Roa-Ureta Loligo gahi stock assessment survey, first season Technical Document, Falkland Islands Fisheries Department. 43

48 Payá, I. 2006a. Loligo gahi stock assessment survey, second season Technical Document, Falkland Islands Fisheries Department. Payá, I. 2006b. Fishing Powers of Loligo Vessels, First fishing Season Technical Document, Falkland Islands Fisheries Department. Payá, I. 2007a. Loligo gahi stock assessment survey, first season Technical Document, Falkland Islands Fisheries Department. Payá, I. 2007b. Loligo gahi stock assessment survey, second season Technical Document, Falkland Islands Fisheries Department. Payá, I. 2008a. Loligo gahi stock assessment survey, first season Technical Document, Falkland Islands Fisheries Department. Payá, I. 2008b. Loligo gahi stock assessment survey, second season Technical Document, Falkland Islands Fisheries Department. Ribeiro Jr P.J., Diggle P.J geor: a package for geostatistical analysis. R-News 1: Roa-Ureta, R Loligo stock assessment survey and biomass projection, second season Technical Document, Falkland Islands Fisheries Department. Roa-Ureta, R. 2005a. Loligo stock assessment survey and biomass projection, First season Technical Document, Falkland Islands Fisheries Department. Roa-Ureta, R. 2005b. Loligo stock assessment survey and biomass projection, Second season Technical Document, Falkland Islands Fisheries Department. Saunders, P.M., Practical conversion of pressure to depth. J. Phys. Oceanogr., 11,