New Mathematical Methods to Help Fisheries Stock Assessment

Lucy Towers
04 March 2015, at 12:00am

FINLAND - By applying new mathematical methods, the uncertainties related to fisheries stock assessments can now be taken into account, thus providing more reliable assessments to support decision-making.

In her doctoral thesis, Henni Pulkkinen, Researcher at the Natural Resources Institute Finland (Luke), explored how the various sources of uncertainty can be taken into account in fisheries stock assessment by using Bayesian statistical models, which enable extensive combining of information.

For example, biological and ecological data on related species can be utilised in the assessment of many endangered and data poor fish stocks.

"Bayesian modelling represents a learning process where existing information is updated with new information. The challenge lies in identifying the wide variety of sources containing useful information. Such information can be obtained from literature or databases, but it can also be so-called tacit knowledge collected through expert interviews," said Ms Pulkkinen, who will defend her thesis at the Faculty of Biological and Environmental Sciences of the University of Helsinki on 27 February 2015.

Nature does not follow a single mathematical formula

While traditional fisheries stocks assessment models are largely designed based on observed data, sufficient attention is not paid to the uncertainties underlying the resulting assumptions. Bayesian modelling enables the description of the whole biological process even if the amount of data available is small. This makes it easier to identify the least known elements and take into account any uncertainties related to them.

In her doctoral thesis, Ms Pulkkinen also discusses model uncertainty (structural uncertainty), which results from the fact that the phenomenon being researched can be explained with several – even contradicting – theories.

"There is not a single mathematical function representing a natural phenomenon that is absolutely correct. However, a combination of several different, even competing views on the best mathematical model can provide a more extensive understanding of a phenomenon than any of them alone. The model choice can have a significant effect on the interpretation of data."

One of the success stories related to Bayesian modelling is the Baltic salmon stock assessment model, which Ms Pulkkinen also reviews in her thesis. The model combines biological background knowledge with extensive research data of the wild salmon stocks in Finland and Sweden. The International Council for the Exploration of the Seas (ICES) uses fisheries stock assessment data in its annual scientific advice concerning the fishing quotas for the Baltic salmon.

According to Ms Pulkkinen, the use of Bayesian statistical models is increasing in population biology. However, modelling requires understanding of probability calculus, and the in the practical implementation of the models computational challenges need to be tackled.

Despite these challenges, Bayesian models have become a daily tool in numerous fields. In fact, most of us use Bayesian applications on a daily basis without knowing it, for example, in search engines, translation software and speech recognition systems.