Modelling viral impact on bacterioplankton in the North Sea using artificial neural networks.

The temporal variability of the viral impact on bacterioplankton during the summer-winter transition in the North Sea was determined and artificial neural networks (ANNs) were developed to predict viral production and the frequency of infected bacterial cells (FIC). Viral production and FIC were estimated using a virus-dilution approach during four cruises in the southern North Sea between July and December 2000 and an additional cruise in June 2001. Supplementary data such as bacterial production, and bacterial and viral abundance were collected to relate changes in FIC and viral production to the dynamics of other biotic parameters. Average viral abundance varied between 4.4 x 10(6) ml(-1) in December and 29.8 x 10(6) ml(-1) in July. Over the seasonal cycle, viral abundance correlated best with bacterial production. Average bacterial abundance varied between 0.5 x 10(6) ml(-1) in December and 1.3 x 10(6) ml(-1) in July. Monthly average values of FIC ranged from 9% in September to 39% in June and the average viral production from 11 x 10(4) ml(-1) h(-1) in December to 35 x 10(4) ml(-1) h(-1) in July. The data set was used to develop ANN-based models of viral production and FIC. Viral production was modelled best using sampling time, and bacterial and viral abundance as input parameters to an ANN with two hidden neurons. Modelling of FIC was performed using bacterial production as an additional input parameter for an ANN with three hidden neurons. The models can be used to simulate viral production and FIC based on regularly recorded and easily obtainable parameters such as bacterial production, bacterial and viral abundance.