Title: A data sampling strategy for coastal herring of Sri Lanka

Herring (Amblygaster sirm) is the most valuable species found in the small scale coastal gillnet fishery of Sri Lanka. The species has a short life span of little over a year and reaches around 20 cm in length. Several studies have been conducted on this species in order to aid sustainable management. Recently conducted studies have revealed that the present level of exploitation of this species in Sri Lanka is unsustainable. Conducting stock assessments using length frequency data and length-based stock assessment models is useful for managing fisheries resources. At present there is no sufficient length frequency data available from present data collection programmes. This was found to be the major obstacle faced when attempting to conduct length-based stock assessment on coastal herring of Sri Lanka. Therefore, attention was paid by this study to reviewing the status of herring and also to proposing a better sampling strategy to obtain length frequency data for future stock assessments. The non-parametric bootstrap technique was used to determine the minimum sample size of length measurements and number of samples required to conduct stock assessment on herring. Capelin (Mallotus villosus) length-frequency survey data obtained from annual capelin surveys in Icelandic waters was used as the proxy population for this analysis. The issue of possible internal correlations in the length measurements in a sample (i.e. only getting small fish in a sample or only large ones) was also addressed while applying the bootstrap method for capelin data. The results from the bootstrapping indicate that around 60 samples, each with around 40 measurements, are needed per month to follow the rapid growth of A. sirm over the year and conduct assessment. The minimum sample size and number of samples are determined by resource expenditure for taking length measurements and the degree of precision of the estimates made from collected length measurements. Therefore, 20 - sample size and 40 – number of samples might perhaps be adequate when cost for data collection is taken into account and this might be the optimal trade-off between cost and parameter precision.