Predicting the temperature of the Barents Sea

Knowledge of the influence of the physical environment on commercially impo rtant fish stocks in the North Atlantic has increased during the last decad e. To allow this information to be used in fisheries management, some forec ast of the environment is important. Predictions of temperature in the Arct o-boreal Barents Sea have been given for many years, both as subjective opi nions of scientists and implicitly in stock assessment assumptions of, e.g. , mortality rates. To evaluate an objective statistical forecasting system, we have analysed time series representing mechanisms previously proposed a s influencing the temperature of the Barents Sea. These include components of suggested periodic nature, large-scale advective effects, regional proce sses, and atmospheric teleconnections. The predictability of Barents Sea te mperature based on the above mechanisms was evaluated through calculations of auto- and cross-correlations, linear regression, spectral analysis and a utoregressive modelling. Forecasts based on periodic fluctuations in temper ature performed poorly. Advection alone did not explain a major part of the variability. The precision of predictions six months ahead varied with sea son; forecasts from spring to autumn had least uncertainty. A first-order a utoregressive model, including modelled atmospherically driven volume flux to the western Barents Sea during the preceding year and the position of th e Gulf Stream off the eastern coast of the USA two years earlier, explained 50% of the total historical temperature variability.