The quality of age determination plays a crucial role in age-structured fish stock assessment. However, as it has been stated in many cases (e.g. ICES, 2009a), consistency in age readings has been very difficult to achieve. Thus, for example the Baltic Fisheries Assessment Working Group of the International Council for the Exploration of the Sea (ICES) realized that there were long-standing problems with age determination inconsistencies in the assessments of Baltic cod and other demersal fish species. The group further concluded that substantial differences occur in age determination for fish from the same subdivision and year. The working group also realized that there were consistent differences between nations ('age reading schools') within and between years, indicating at different interpretation of growth zones on otoliths (ICES, 2009a). Regular exchange of otoliths and arrangement of workshops of age readers have been proposed (and proved to be useful) in order to improve the situation (ICES, 2008, 2009b). However, these costly measures can be effective only if taken regularly. Therefore, it comes as no surprise that looking for alternative methods to distinguish between cohorts (year classes) in analytically assessed fish stocks has become increasingly topical (e.g. Cardinale et al., 2000).
Virtual population analysis (VPA, e.g. Lassen & Medley, 2001) is a widely used method of age-structured stock assessment, deployed also by the ICES in its annual assessments. Stock assessment in the VPA is an analysis of the catches of commercial fisheries, obtained through fishery statistics, combined with detailed information of each age group (cohort) in the catch, which is obtained through sampling programmes and age readings. The idea behind the method is to back-calculate the population composition that must have been in the water to produce this catch (Sparre & Venema, 1992).
The counting of seasonal growth differences expressed on the otoliths as 'winter and summer rings' has been the most widely used method for age determination (ICES, 2008). In addition, analysis of otolith weight distribution (Pawson, 1990; Cardinale et al., 2000; Pilling et al., 2003; Pino et al., 2004; Cardinale & Arrhenius, 2004) and analysis of total length distribution of catches have been used to discriminate between the cohorts (Vitins, 1989; Sparre & Venema, 1992).
There are several techniques for age reading using otoliths, which depend on the species. Generally the method of stained slices is considered the most precise, and also burned and broken otoliths are considered to be better readable than whole otoliths (ICES, 2008). However, the more exact age reading methods (e.g. the method of stained slices) are also more time-consuming and expensive because special laboratory equipment is needed. As a result, the number of analysed specimens usually decreases.
The annual rings of otoliths of younger age groups are better readable than in older fish (ages 7 years and more). As the relatively young flounder (age groups 3-5) are dominating in the Estonian commercial catches, the counting of annual rings can be done effectively using the whole otolith (no slicing or breaking).
Several recent studies show that the weight of otoliths can be used for age determination (Pawson, 1990; Cardinale et al., 2000; Pilling et al., 2003; Pino et al., 2004; Cardinale & Arrhenius, 2004). Therefore, using different age determination methods simultaneously can be valuable in order to verify the age readings. In this respect, implementation of proper methods and criteria to discriminate different cohorts in the weight or length distribution is of crucial importance.
Bhattacharya's method, distinguishing parameters with normal distribution from a combined distribution pattern (Sparre & Venema, 1992), can be used for splitting composite distributions of otolith weight and fish length distributions into separate normal distributions, applicable to cohorts. …