Traditional trawl-surveys are characterized by a standardized methodology regarding the distribution of hauls, duration of tows, use of a unique fishing gear and single fishing procedure, etc. This fact have a great influence on the qualitative and quantitative characteristics of the catch and constitutes the main reason why seldom the species composition of catch and the size structure of all the involved species is similar to those obtained by the commercial fleet. In the other hand, if for a certain species all size classes are completely vulnerable to the utilized gear, the trawl-survey data will give an unbiased representation of the real demographic structure of the surviving population. If this is also true for the majority of the other species, trawl-surveys will give also a better information on the relative importance (in biomass or number) of each one of these species at sea. Commercial vessels look for higher concentrations of high-prized species and discard of less valuable species (or size-classes) is quite common. In consequence, the relative weight of each species in the landings will not necessarily correspond to their relative importance at sea. Size frequencies of commercial landings not necessarily reflect the demographic structure of the stocks at sea because effort will be allocated to the best fishing areas (but this doesn't mean that from these data we will never be able to reconstruct the true stock structure at sea)

If we are interested to assess the state of a fishery we need an instrument that allows the modelization of the fish population reactions to different management actions as well as the definition of a sustainable production or level of fishing pressure over time. It is evident that a knowledge of the commercial fishing strategies, of their changes in space and time and of the impact of each one of them on the stocks will be absolutely necessary for this goal. In the Mediterranean, however, commercial data collection is very difficult because many species are involved, many fishing strategies are utilized, there is an important (and difficult to quantify) discard at sea of many species as well as of certain size classes of commercial species. The same species are caught with many gears, landings spread out along the coast line, many times fish boxes are composed by individuals of several species or commercial categories of a single species and in consequence, the obtaining of representative samples of the structure by size of the catch is very difficult and expensive. For these and other reasons it is quite difficult to organize efficient systems of catch, effort and size structure data collection and often statistics are incomplete. In consequence, the estimation of total amount of catch, total effort and its partitioning, size structure of catch, etc. are seldom precise and accurate. Even for the countries where a system has been implemented, long time series are rarely available.

Notwithstanding that a correct stock assessment needs a detailed knowledge of the fisheries, we will try to demonstrate that some times (and for some species) a preliminary stock assessment mainly based on trawl-surveys data is possible to attempt with both analytical and surplus production approaches. This note does not pretend to be exhaustive regarding the several possibilities of trawl-surveys data analysis for stock assessment purposes.

The classical " structural " model of Yield-per-Recruit (Beverton & Holt, 1957) has been widely utilized for stock assessment in the Mediterranean. It is a " parameters-based " model and data necessary for their estimation may proceed from different sources. As soon as all the necessary parameters are available, we can estimate the yield produced by any change in fishing effort and age of first capture. Classical Von Bertalanffy Growth parameters K, L¥ and t₀, Length/Weight relationship parameters a and b and an estimate of natural mortality constitute the basic input. By means of the trawl-surveys we can collect data (length frequencies, individual weight), analyse them and quite easily to estimate the required parameters. In traditional versions of yield-per-recruit analysis, natural mortality rates are considered constant beyond the age at first capture. M can be estimated utilizing some of the available empirical equations as Pauly (1980), Rickhter & Efanov (1978), Hoenig (1986) based on information regarding growth performance, species longevity, etc.

For the estimation of the current fishing mortality rate F, if a complete catch assessment survey has not been performed, we have no available size or age structures of the commercial catch and in consequence, we need to analyse the structure at sea proceeding from trawl-surveys data. First at all, we have to do some assumptions in order to make this alternative information usable. The main necessary assumption will be that the structure by length or age so obtained is representative of the average demographic structure of the population at sea and in consequence, the declining in numbers observed in the right part of the so called " catch curve " will be due exclusively to mortality. By means the utilization of regression analysis techniques applied to information corresponding to the survivals for each length (age) interval we can estimate the instantaneous total mortality rate Z and then to calculate F as a difference between Z and M.

A " catch curve " that represents the mean annual structure of the population can be obtained by merging weighted size distributions of the total catch derived from several surveys performed during different periods of the year.

When we are dealing with length converted catch curves, the choice of the parameters of the chosen growth model is very critical because these parameters will allow to assign a relative age to each length class and to estimate the time necessary to grow from a certain length class to another. When growth performance for both sexes is very different, for the analysis of data, the utilization of a unique set of Von Bertalanffy growth parameters without a distinction by sex may produce a serious overestimation of the total mortality rate. Moreover, considering that the utilized technique is a simple linear regression analysis, we have to assume for the whole analysed portion of the curve an equal instantaneous rate of decrease. This means, for the whole exploited phase, equal fishing mortality rate F and also equal natural mortality rate M. If the gear selectivity determines a very small age of first capture, a similar natural mortality for all the involved age classes is not hypothizable because the younger individuals are likely to show a higher mortality rate. Due to the fact that in the Mediterranean for many species recruitment to the fishery starts too early in their life, the suitability of this method for the estimation of Z will be so restricted to a few species. Moreover, for species with very high natural mortality and short lifespan, the usefulness and precision of the analysis of the catch curve technique is reduced because the " catch curves " due to problems related to the recruitment strength and to the nature of size distribution of the different cohorts that constitutes the exploited fraction of the population.

Estimators based on mean length of capture like Beverton & Holt (1956), Hoenig (1987), etc., have been frequently utilized for a rough estimate of Z. However, they present the same problems above described for the LCCC technique and trawl-surveys data generally do not allow to estimate the mean size of the catch necessary for the estimation of Z with these methods.

Other approaches, traditionally applied to commercial catch at age data, are the so called "Virtual population analysis" and similar techniques. With these recursive methods we can estimate not a single fishing mortality rate but a vector of F as well as the number by age (or size) at sea. It is evident that the precision of results will depend on the quality of input data (total catch number at age or size and biological parameters). For the species exclusively caught with bottom trawl nets, if the trawl-survey catch structure by size is representative of the structure of the landings, in the best of the hypothesis, this technique will allow to estimate an F vector that in theory should be identical to the obtained with commercial data. If applied to trawl-surveys, due to the nature of data, the method will never allow to the estimation of stock size. Traditional versions of retrospective analysis (VPA, LCA, etc.) do not enables the utilization of a vector of M at age.

The simulation of different scenarios and the assessment of their consequences related to yields are however only the first step aimed at the species stock assessment. We have to define, among many possible ones, which is our management objective (maximize short or long term yields, maximize future recruitment, maximize economic efficiency, avoid recruitment overfishing, etc.). It will be necessary to find some reference point (expressed in terms of fishing effort, fishing mortality rates, biomass levels or total annual catch) that is expected to ft our management goal. One of the most important aspects that we have to consider is the sustainability of the chosen fishing strategy that we have considered as the optimal. However, seldom we will be able to estimate the Maximum Sustainable Yield or levels of effort or fishing mortality corresponding to this optimal yield level. (n consequence, we have to choose another reference point based on considerations other than the MSY concept, that likely will furnish a yield close to the MSY.

If we choose a Yield-per-Recruit based approach, frequently we will observe that the yield-per-recruit isoplets over a certain level of F are very flat. In these situations, beyond a certain F level, a small increase of Y/R should be obtained with a high increase in effort (and in fishing mortality). Fmax is in these situations a very bad target reference point because with a high effort increase the catch rates and the economical efficiency of harvesting will be reduced while the stock biomass will be extremely small and so the stock reproductive reserves.

Different combinations of fishing rate and age of first capture ft any one of the above mentioned management goals but the combination that allows to reach certain goal could have very negative consequences regarding others. The F_0.1 criteria proposed by Gulland & Boerema (1973) seems a quite conservative reference point and often constitutes a good compromise for the majority of the species. It usually provides good yields near the MSY and it is not so likely to deplete the spawning stock. It has been adopted by many countries for the formulation of management advice. Another advantage of this reference point is that it is quite easy to estimate.

On the last years, reproduction-based reference points (F% SSB, F0.5, eggs per recruit) have been more frequently used. The utilization of reproductive constraints as well as other complementary data sources (as commercial data) should improve the results that will be obtained.

Mediterranean fisheries are very diversified and dynamic, many species and fishing gears are involved and important changes in time and space do occur. How to give advice for these multispecific stocks? Is it possible to estimate an F value in order to guaranty the self-renewal for all the species that constitute a certain multispecifc assemblage exploited with one or more fishing strategies.

MEDITS data are not suitable for the definition of the demersal commercial assemblages. In fact, the assemblages defined through the analysis of data proceeding from the trawl-surveys standardized procedures (equal duration of hauls, random allocation of fishing stations, work only during the day-hours) do not coincide with those defined by the analysis of the commercial fishing strategies (characterized by heterogeneity of gears, fishing procedures, geographical distribution of effort, with seasonal changes in target species mainly dependent of market constraints, etc.

MEDITS data look, at a first sight, not suitable for their utilization in Surplus Production Models. For traditional versions of these approaches, long data series of catch and effort are needed. Moreover, the available data on effort levels and corresponding catch rates have to be " contrasting " (catch rates produced in different conditions of moderate, medium and excessive fishing pressure) in order to obtain a good fitting of data. However, many of the problems above mentioned can be avoided if some variants of the Production Models are chosen. For instance, Caddy & Csirke have proposed a production model that utilizes Z, Z/K or F as direct indexes of effort and an abundance index (catch rates as CPUE or Y/F). The main problem that still remains is the need of long time series (MEDITS data series is only three years long !). In order to avoid most of the last mentioned problem, spatial information proceeding from several subareas exploited with different rates can replace temporal series (the so called " Composite Models ") as proposed by Munro (1980). We only need to assume for all the areas included in the analysis similar initial productivity and evolution under any change in fishing pressure. This assumption is quite difficult to verify but it is probably not less realistic than certain assumptions necessary for the application of the traditional approach with time series. Commercial catch rates, in fact, are not always a good index of abundance at sea. Considering that fleet movements are not random, for some schooling fish, declining in catch rates or in other indices of fishing impact could not be observed because the schooling behaviour assures high density even when total abundance is already reduced. The efficiency of fishing fleets also changes in time and in consequence also the relative catch efficiency per unit of effort. Inside the traditional approaches the equilibrium assumption is implicit and this assumption in seldom true. In the other hand, the Munro approach assumes that each area exploited with certain rate is at equilibrium respect to its effort. This is probably the more frequent situation in the Mediterranean fisheries (at least in Italy) because in the last years most of the fisheries did not show important changes in effort for each area.

Another advantage in utilizing Z instead of effort in a " composite model " is that we do not need to consider the fishing intensity (effort per unit area) and standardize the catch rates as Garcia did for a stock assessment of demersal species in the Gulf of Lyons. He utilized catch per unit effort and effort data from different areas exploited with different rates.

For the utilization of a " composite model " first we need to define homogeneous areas, each one of them characterized by a given fishing pressure. Complementary data from Commercial Catch Assessment Surveys might be utilized for this definition. For each one of the so defined areas a mean annual catch rate and the mean annual total mortality rate Z have to be estimated. If data from different areas proceed from catches of different vessels a standardization is necessary in order to make data comparable. For the estimation of Z, the same technique as well as same biological parameters must be utilized for all the areas included in the analysis. The utilization of couples of catch rates and Z proceeding from trawl surveys never allow to estimate the Maximum Sustainable Yield but we can estimate the so called Maximum Biological Production (MBP) that is defined as the maximum production that can be harvested by fishermen or removed by natural mortality and should permit us to know which is the current situation of each species or for a group of similar species in relation to the MBP. This approach is appropriate even if M and growth parameters are unknown. In fact, by means of the analysis of the data on size structure proceeding from a random stratified survey (Medits) with the Powell-Wetherall or the Jones & Van Zalingè methods, we can estimate Z/K. Being K a constant value, we can utilize Z/K instead of Z obtaining the same results.