No abstract

Time-frequency analysis (TFA) has been widely used in seismic processing and interpretation. A good time-frequency representation can preferably characterize geologic spatial distribution and detect hydrocarbon reservoir anomalies. This paper applies a robust seismic TFA method based on the general linear chirplet transform (GLCT). The GLCT method is an extended form of LCT, which is a unifying framework encompassing the short time Fourier transform (STFT) and the continuous wavelet transform (CWT) using the chirplet atom as the kernel function instead of the sinusoidal wave or wavelets. By rotating the chirplet atom at each time-frequency point, GLCT method could adaptively choose the best atom to fit the local time-frequency feature of seismic signals. The algorithm follows such a simple logic and produces a broadband time-frequency spectrum free of cross-term interference, resulting in good performance characterizing the instantaneous spectral variations. Synthetic data analysis demonstrates that the GLCT method is able to reach a higher energy concentration in the time-frequency plane than conventional methods. Robustness analysis indicates that GLCT produces more stable results that outperform not only STFT, CWT, but also high-resolution methods such as the synchrosqueezing transform and complete ensemble empirical mode decomposition in the case of noisy data. The application to field data illustrates that the isofrequency attributes extracted by GLCT through spectral decomposition could effectively image subtle stratigraphic structures of the subsurface paleotopography and highlight the frequency anomalies associated with hydrocarbons. Sometimes, these anomalies might be otherwise inundated in the background noise. Our method can be a validation tool for seismic facies interpretation improvement and direct hydrocarbon indication in practice.

Temporal changes of seismic velocities in the Earth’s crust can be induced by stress perturbations or material damage from reasons such as strong ground motion, volcanic activities, and atmospheric effects. However, monitoring the temporal changes remains challenging, because most of them generally exist in small travel-time differences of seismic data. Here, we present an excellent case of daily variations of the subsurface structure detected using a large-volume air-gun source array of one-month experiment in Binchuan, Yunnan, southwestern China. The seismic data were recorded by 12 stations within ∼10 km away from the source and used to detect velocity change in the crust using the deconvolution method and sliding window cross-correlation method, which can eliminate the “intercept” error when cutting the air-gun signals and get the real subsurface variations. Furthermore, the multichannel singular spectral analysis method is used to separate the daily change (∼1 cycle per day) from the “long-period” change (<1 cycle per day) or noise. The result suggests that the daily velocity changes at the two nearest stations, 53277 (offset ∼700 m) and 53278 (offset ∼2.3 km), are well correlated with air temperature variation with a time lag of 5.0 ± 1.5 hr, which reflects that the velocity variations at the subsurface are likely attributed to thermoelastic strain. In contrast, both daily and long-period velocity changes at distant stations correlate better with the varying air pressure than the temperature, indicating that the velocity variations at deeper depth are dominated by the elastic loading of air pressure. Our results demonstrate that the air-gun source is a powerful tool to detect the velocity variation of the shallow crust media.

Active sources, especially air-gun sources in the water reservoir, have proven to be powerful tools for detecting regional scale velocity changes. However, the water level change affects the repeatability of the air-gun waveform and, thus, affects the stability of the detection of the velocity changes. This article explores how to make full use of the air-gun signals excited at different water levels from analyzing three years of air-gun data recorded by 20 stations deployed from ∼50 m to ∼25 km from the source. At the same time, by utilizing the poroelastic model, we quantify both vertical and horizontal distances affected by the water level change. More important, supported by the strain data from one borehole strainmeter station, the influence mechanisms of the seasonal variation derived from the three years of air-gun data are also discussed. Results indicate the water level affects the main frequency of the air gun and has a strong influence on the coda wave. When the water level of the reservoir changes abruptly, the dominant effect on the derived delay time change is from the water level change. In this case, the deconvolution method can hardly eliminate the influence of the abrupt water level change. However, when the reservoir's water level changes gently, the delay time varies accordingly rather than inversely with the water level. Other reasons affect the material properties and, thus, influence the derived delay time. The modeled vertical component of poroelastic strain caused by the reservoir water level change is 1×10−7. The observed strain (4×10−7) from the strainmeter is likely associated with thermoelastic strain induced by temperature change. Our results show that although the long-term air-gun signal is affected by water level, there is still much information about changes in the subsurface that is worth mining.