This paper investigates the use of spectral decomposition for extracting information on fluid properties. Traditional theory for detecting fluid response is based on the pure elastic Gassman theory, and the resultant seismic effects are frequency-independent. Using dynamic fluid substitution, we demonstrate that the frequency response of seismic reflection and its resultant attenuation and dispersion are directly linked to fluid saturation. To extract this information, we develop an accurate two-stage spectral decomposition method by matching pursuit. This allows us to calculate a range of frequency-dependent attributes, such as, absorption coefficient and amplitude gradient in the frequency domain. Application to real data shows a good link between the anomalies and hydrocarbon saturation. The results highlight that careful data processing and modeling are necessary to understand the complex effect of different fluids on the spectral response and enable robust interpretation.