Assumed that climate change is a period signal only changed with time, and a composition of multi-timescale signals without regard to the affect of other physical factors, such as sunspots, radiation etc., then each periods of climatic data can be decomposed by EMD method and the model can be built using nonlinear minimization procedures to reconstruct this climate data.Further, the future climate change can be estimated by using this model.The experimentations on two different real time series, global sun solar in January and North hemispherical temperature anomalies data from NOAA.CDC., indicate that this method is able to get very good fit results.Therefore, the method used in this paper show great promise for revealing patterns in historical data.