CART (Classification And Regression Tree) is an effective tool to study high dimensioanl, complex data sets.  In this research, we propose a method for scenario generation based on one historical trace using the CART methodology.  The method can apply to not only short-range dependent (SRD) data, but also long-range dependent (LRD) data.  We show its advantages over the existing ones using some new and traditional criteria.  Its applications in computer network traffic modeling, Monte Carlo simulation, and automatic generation of lifelike scenarios in the field of finance are also explored.  In addition, when applying this methodology to LRD data, the distribution pattern of the complexity parameters behaves substantially different from that of SRD data.  Such difference provides some potential new approaches to distinguish LRD phenomena and to study the corresponding data sets.  Finally, we show the theoretical justification of our method in the wavelet domain.