Automated modelling of a food web in lake Bled using measured data and a library of domain knowledge

In this paper, we applied automated modelling (computer model construction) method to the task of modelling a complex lake ecosystem. The method (Lagramge) integrates domain expert knowledge in the process of automated model induction from given data set. The data set comprises long-term measurements (from 1995 to 2002) of physical, chemical and biological data in lake Bled, Slovenia. Given expert knowledge in terms of a simple food web concept and rules for modelling thereof, we first induced a model for long-term dynamics of the phytoplankton in the lake. Failing to obtain a good fit, we also induced models of phytoplankton dynamics for each year separately. The differences between these models indicate structural dynamics of the food web in lake Bled, i.e., indicate that the behaviour of the lake is changing from year to year. Additionally, we successfully induced a three-equation model (nutrient-phytoplankton-zooplankton) on the data from year 1996.