This study focusses on realistic modelling of gravitational wave interferometric data. We estimate the non-stationarity of the noise floor and model it using linear timeseries models. A robust line tracking algorithm is used to estimate the amplitude and phase of all major lines in the data and model them. Features of the modelled data are discussed. These models are used to generate realistic data that share important characteristics with the data used for training the models. In this way we can generate long stretches of "playground" data which could mimic more truly the behaviour of real data at any one instant than long stretches of real data used as playground. Examples are drawn from LIGO S2 data. Importance of generating realistic simulated data and its applicability in connection with astrophysical searches are also discussed.