In my project I explored the capabilities of the Ensemble Kalman Filter for the detection of model errors in the context of ecosystem dynamics. To this end I modeled small food webs with a generalised Lotka-Volterra model and investigated scenarios with different types of model errors.

The importance of understanding ecosystem dynamics

Mankind benefits from many ecosystem functions and services, that range from the production of food and raw materials, to waste treatment and climate regulation. The extent to which these services can be utilized depends critically on the dynamics of complex ecological communities. Because of this, understanding and quantifying the processes driving the community dynamics is necessary to predict the degree to which the ecosystem services can be harnessed without causing critical transitions in the system. An example of this is the amount of fish that can be extracted from an ecosystem, without decreasing the long term availability of fish stock.


The following figure shows a food web with 10 species and the mean flow of biomass between them.

State-of-the-art ecosystem models can have thousands of parameters that are expensive to determine in experiments. Therefore it is important to determine which processes are relevant in a specific ecosystem


I have found that the quality of the estimated parameters depends largely on how well the influence of the parameters can be separated on a short time scale.