BackgroundFishtalk is a software product used in production control for farming fish. Biological, financial and environmental information is recorded, reported and analysed. Fishtalk uses several models to simulate growth based on a number of factors, like feed consumption and environmental measurements.
The harvest size distribution is often presented as a normal distribution. At harvest time the CV (Coefficient of Variance) can be in the range of 18-25%, and this variation has a huge impact on the actual product to deliver, which is typically boxes of fish in intervals of 1 kg. Orders are placed on specific sizes, so if the distribution is far different from the estimate, the producers end up with too much fish in some weight intervals, and too little in others. This makes the distribution and sales operation more expensive, as alternate solutions must be found in short time, typically on the day of harvest.
This project is aiming to add size and quality distribution estimation to Fishtalk by means of models using a large database of harvested fish groups as input. We are also aiming at finding markers in the data set that indicates when there is an increase in possibility that there will be a large deviation between estimated and actual stock in the fish cage.
MethodsThe project has used different statistical methods for identifying parameters or factors influencing the stock status and the size distribution.
Among the methods used are the ASMOD framework developed at SINTEF. The ASMOD (Adaptive Spline Modelling of Observation Data) algorithm is an algorithm for empirical modelling. It uses B-splines to represent general nonlinear and coupled dependencies in multivariable observation data. The model is iterative to find the best parameter representation of the data. Methods like student’s t-test and ANOVA (analysis of variances) have been used to test hypotheses regarding features initially believed to have a significant influence on the fish population in the cages.
Results and discussionThe majority of work so far has been based on data from the 2006 generation. We see that the data verify the biological theories on growth like the importance of temperature and feeding.
We also have identified that the size distribution seems to have:
Published July 6, 2009
Eivind Brendryen (AKVA group)
AKVA group ASASINTEF Information and communication technologyNOFIMA Marin