Carbon and latent heat fluxes can be simulated with different model strategies to fulfil different research purposes. In this study we compared four different model concepts: artificial neural networks (ANN), fuzzy logic (FL), an index model (IM, using light use efficiency and water use efficiency) and the process based model FORGRO. The models were tested on a 2-year data set of carbon and water fluxes of a Douglas-fir forest, 1 year before and 1 year after a thinning. The potentials of the model concepts for application for four research goals were assessed in relation to the obtained results and in a more general context: measurement fitting, insight into the importance of processes and mechanisms, simulation of climate change effects and up-scaling of forest responses to regional scale. For measurement fitting ANN and FL showed the highest potentials, mainly because of their high number of fitting parameters. IM and FORGRO showed a satisfactory model performance, although systematic errors were detectable. Insight into forest ecosystem functioning was difficult with ANN, but FL, IM and FORGRO showed clear interpretability of the effects of the thinning in terms of ecosystem functioning. FORGRO has the highest potentials for reliable estimation of effects of climate change on forests like Speuld, although the incorporation of adaptation to climate change in the model formulation is a major problem unsolved. For up-scaling FL and ANN can be used effectively if they are parameterised on a range of forests rather than one forest as in this study. IM showed potentials for linking the model parameters to variables characterising forest ecosystems like leaf area index, and thereby for large-scale applications. The discussion showed that the application of a set of totally different models can increase our knowledge of forest functioning.