Abstract: Data assimilation is a process where data from two unprecise data sources are used to produce a better prediction. The assimilation process is obtained from a weighted combination between incomplete and inexact experimental with inaccurate mathematical model data. Some schemes to perform the data assimilation will be presented; starting from classical methods, sucg as nudging and optimal interpolations techniques, up to moderns approaches: variational formulations, Kalman filter, and artificial neural networks. Some examples using non-linear models are worked to illutrate data assimilation process.