Improved predictive models of growth, production and stock quality under different habitat scenarios

SEAwise research

The ecological factors studied in this research varied by Case Study region. The main biological processes examined were: growth (both weight and length) and size at maturity, while the ecological drivers include: temperature, salinity, prey abundance, population size and fishing mortality. To assess the predictive capabilities of the model, a hindcasting approach (whereby past conditions are simulated to allow for comparison of actual observations) was used where the data set is split into a training dataset and a test dataset, and the results compared. 

• For the Gulf of Riga spawning herring in the Baltic Sea, the impact of climate, food and density-dependence on growth was studied. The explanatory variables considered were spawning stock biomass (i.e. the amount of mature fish in weight), temperature and prey abundance. The data was split into two time periods: 1961-1988 and 1989-2020. In general, the significant drivers differed in both periods and the explanatory power of the model for the second period (1989-2020) was better than for the first period (1961-1988).
• In the Mediterranean Sea, our analysis focused on effects on size at first maturity, condition factor (i.e. an estimate of well-being) and growth. In most cases, the most significant environmental driver was bottom temperature (i.e. the temperature at the ocean point where the water meets the seabed), although some relationships with bottom salinity, primary production, zooplankton and macrobenthos (AKA organisms that live on the bottom of the water column and are visible to the naked eye) were also found. 
• In the North Sea, the performance of a variety of models were assessed in terms of model fit and predictive capability for seven stocks. Cod, saithe, haddock and whiting showed forecast potential for at least one of the methods examined. In addition, otolith increment analysis was used to study the development of growth of sole in the North Sea and in the Irish Sea.
• In Western Waters, the first study analysed length at age of four species based on Irish groundfish survey data. Across areas, temperature had a negative effect on all age groups for whiting, haddock, and megrim, with more variable impacts observed in blue whiting. In the second study, a variety of models were applied to sixteen Western Waters’ stocks. In the Irish Sea, four out of five stocks showed forecast potential, while in the Celtic Sea three out of six stocks showed forecast potential. The third study in the Western Waters analysed relative condition factor in the Bay of Biscay and the Celtic Sea and found a significant decrease over time for 12 species, while two species showed an increase, and three species showed no significant trends. Finally, a temperature dependent growth model for European seabass showed higher growth increments at higher temperature, implying a faster growth for seabass in the Bay of Biscay, compared to the English Channel, Celtic Sea and North Sea.
• Beyond the regional case studies, a range of experimental studies were conducted to analyse the impact on European seabass of changing environments. The impact of ocean warming and acidification on growth, endocrine, intestinal, and metabolic parameters, of ocean deoxygenation on metabolic rates, of ocean acidification on sexual maturation and of contaminants and pollutants on bioaccumulation, growth and physiological functions were studied.

The ability to predict future events (AKA predictive ability) is one of the most important features of a model and was evaluated for some models using a hindcasting approach. The accuracy of the model predictions was measured in terms of average differences between predictions and observations. Overall, the results indicated that the best models in terms of model fit (or how well they fit the data) did not always have good forecast skills.