Spatial distribution pattern generating processes in the International Bottom Trawl Survey in the,North Sea

Abundance indices ofyoung fish in the North Sea derived from the International Bottom Trawl Survey (JBTS) show large interannual fluctuations, which are not in accordance with ICES assessments in every case. The standard estimation of the abundance indices, however, does not explicitly account for potential influences of environmental variables on the spatial aggregation of juvenile fish and the probability distribution of the catches. The present study aims at improving the analysis and tbe estimation procedures for IBTS abundance indices. Variability of bottom trawl catches of age l and 2 whiting and cod was . studied on three spatial scales: at fixed stations, within statisticaI rectangles and within the species-specific standard areas used in the IBTS. Daylight period, temperature, salinity, water depth. and bottom type were considered as potential environmental variables. influencing the catches. At fixed stations significant diff~rences between day and night catches were found .. Variability within rectangles increased with the degree of heterogeneity in theenvironmerit, where the effects of temperature and depth were most pronounced. Frequency distributions ofthe average catch by rectangle always deviate substantially from normal distriputions. Lognormal-based Generalized Linear Models for whiting and Delta-Iognormal models for cod revealed that temperature and, in particular, depth contribute significantly to the variability also between rectangles. It is further demonstrated that depth can be used as a covariate in the estimation of abundance indices for the species-specific standard areas. Introduction The International Bottom Trawl Survey (JBTS) in the North Sea conducted in February each year started in the beginning of the 1960's and aimed in the start exclusively at juvenile herring. Over the years the objectives of this survey were broadened to provide recruitment indices also for cod Gadus morhua, haddock Melanogrammus aeglejinus and whiting Merlangius merlangus, and the survey area was extended in the 1980's to include the northern North Sea and the SkagerraklKattegat (Heessen et al. 1997). At the same time the GOV (Grand Ouverture Verticale) trawl was introduced as standard gear and in the early 1990's a large amount of effort was put into standardizing the sampling'design (ICES 1992, 1996a). The survey is stratified based on a grid of statisticaI rectangles ("squares") of l o longitude and 0.5 o latitude (~30 ~ 30 nm; Fig. 1). Ships oftwo different countries usually fish each 'square so that at least two hauls are made per square, and the tows conducted 'fromone vessel in adjacent squares are normally separated by at least 10 nm. The survey provides abundance indices that are calculated by taking the average catch per hour trawling of all hauls within a square, and then the average of all squares within species-specific standard areas. Although it is recommended that trawling should be Iimited to daylight hours, some hauls are made during twilight or night due to a short daylight period, and night catches are only excluded from the index calculation for herring (Heessen et al. 1997). The fIrst quarter IBTS abundance indices show large interannual fluctuations, which do not coincide with VPA estimates in every case. For gadoids where multiple survey information isavailable, the fIrst quarter IBTS performs less well for cod than other surveys (ICES 1996b, Wieland 1998), and there is an inconsistency between the surveys and the assessment for whiting in general (Cook 1997, ICES 1998). However, the present procedure of calculating the abundance indices does not consider the error structure of the catches. The use of a rigorous statisticaI approach can potentially improve the perfoimance ofthe survey indices (e.g. Pennington 1983, Stefånsson 1996). There are numerous technical aspects of trawling which affect survey catches, and similarly the behaviour and the distribution of the fish in relation to environmental conditions influence catch efficiency. For example, diumal variations in bottom trawl catches have been observed for different fish species includmg gadoids (Ehrich & Groger 1989, Engås & Soldal 1992, Michalsen et al. 1996), and interannual differences in catchability due to changes in the distribution related to water temperature and salinity were found for cod (Smith & Page 1996). In the present study, the variability ofbottom trawl catches for age l and 2 whiting and cod is analyzed on three spatial scales with different aspects. These are: l) the effect of the daylight period on catches from fixed stations, 2) influences rif environmental variables on catches obtained within a statisticaI square, and 3) the application of statisticaI methods which account for the frequency distribution of the catches, and which also consider environmerital variables as cov:ariates in the calculation of abundance indices for the IBTS species-specific standard areas. Material and Methods Variability atfIXed stations Basic data for the analysis ofvariability ofbottom trawl catches at fixed locations were available from investigations carried out by the Danish Institute for Fisheries Research. The surveys were originally designed to sample whiting stomachs (riegnbol 1992, Pedersen 1996, Pedersen submitted). The sampling was conducted in autumn 1988 and 1989 and in spring 1992 and 1993 with RN Dana. Hauls were made in about 4 hour time intervals over a 72-hour period each at a single station. In 1988 and 1989 the tracks of the hauls were laid out in a star pattem crossing through the same center, and are thus only exact replicates conceming their central position. In 1992 and 1993 parallel tracks close to each other in a box of 5 * 7 nm were used. In some cases fishing w as' conducted at some distance from , ." the bott()m d\lring l'light,. ~ut these tO\vs were discardeØ frQm mIr allalysis. In total 93 hauls were selected for analysis which were all carried out with an EXPO trawl. Sampling periods and statisticai rectangles (Fig. l) in which the sampling was performed are given in Table l. Towing time was 60 min in most cases, towing speed amounted to 3.4 kn on average, and vertical net openingranged between 5 and IO m.