The dynamic Allan variance (DAVAR) is an effective method for analyzing non-stationary signal. However, it has defects such as power leakage and single quantification in noise identification. Therefore, window function combination method and two-dimensional expression of noise value are introduced for their improvements. They are used for analysis and quantitative measurement of various noise terms in the fiber optic gyroscope dynamic error. Based on the dynamic error resolution of fiber optic gyroscope, rectangular window and Hanning window are applied to analyze low and intermediate frequency noise and high-frequency noise, respectively. According to the principle of DAVAR, the noise variation laws with the numbers of sampling point are obtained, namely, the two-dimensional expression of noise value. The experimental results indicate that window function combination method satisfies the identification requirements of noise in different frequency ranges and reduces power leakage; moreover, the changing characteristics of every noise item in the dynamic error are accurately reflected by the two-dimensional expression of noise value.