In order to assess the efficiency of eutrophication control programs, predictive models are necessary. We propose a methodology for implementing such a model, based on the coupling of a biogeochemical model of a lake and the use of long term time series of meteorological data. This methodology is applied to lake Bourget (Savoie, France). It allows to obtain both mean and standard deviation (first and second order moments) of the state variables of the model on a 1 y period. The sensitivity of the model to the various forcing variables, as well as to the initial conditions is analyzed as well as the linear or non-linear behavior of the model. Finally, the propagation of the uncertainties (standard deviations) in time and space, for both water temperature and dissolved oxygen are assessed.