Statistical modelling of temperature variability in the Barents Sea

Abstract

During the latter years an effort has been made to find out more about the relations between environmental variation and recruitment, growth, distribution and migration of fish. The rationale has to a large degree been the needs from fisheries management. To utilize this knowledge for management purposes it is necessary to be able to make some kind of forecast of the environmental situation. This work is an early attempt to quantify the future temperature development in the Barents Sea. We use three different methods, all applied to the ocean temperature time series from the Russian Kola-section. The first method uses the principle of least squares to fit a sum of Fourier components to the observations and construct a function which generates future values. We also apply Holt- Winters models with a linear trend and either an additive or a multiplicative seasonal component. The third procedure classifies the different years into a few categories according to temperature. Statistics on the historical temperature patterns can then be used for forecast purposes. Our results indicate temperature conditions below the long term mean up to 1999. The uncertainty of the forecasts grows with the time-span, but we believe that the picture for the 2-3 first years is reasonably reliable. Our hope is that this work will help towards taking the environmental situation into consideration when evaluating the future fisheries resource situation.