Multi-scale analysis is the main method in identifying landscape spatial characteristics, but lack of classification and comparison evaluation concerning so many different methods used in present ecological researches. These multi-scale analysis methods were categorized based on spatial numerical variable and type variable. In order to compare and evaluate these methods, taking Three-North shelter forest as a study case, the commonly used methods were implemented to identify the characteristic scale. The results showed that the wavelet had a reasonable and clear process, and its outcome measures were time saving. The semivariance was much flexible and its results were clear but consuming time. The outcome of scale variance and lacunarity for characteristic scale detecting were almost indiscernible. We concluded that the most optimal way of scale analysis lied in the combination of semivariance and wavelet. Since the former’s theory and application is simple and robust but can not provide the holistic understanding which can be made up by the latter.