Time-frequency analysis (TFA) can reveal geologic information from seismic reflection data and has been widely used in characterizing geologic bodies, reservoir identification, and gas detection. To improve the energy concentration of TFA, we have developed an extension of the W transform (WT) which incorporates a time-varying Gaussian window that adaptively changes according to the characteristics of the seismic data. The new transform addresses a perceived defect in the original WT where the time-frequency amplitude peak can split at the dominant frequency of the signal. The three parameters of the transform are chosen through an optimization procedure that relies on the concentration measurement and Rényi entropy. We illustrate that our transform overcomes the dominant frequency splitting artifact of the WT and improves the time-frequency resolution by using synthetic data and field data from the Zhongjiang gas field in western Sichuan, China.