GEOPHYSICAL METHODS

Passive low frequency spectral analysis: Exploring a new field in geophysics

This pathfinder/DHI technique is under rapid development and uptake for exploration.

Ren� Graf, Spectraseis Technologie AG, Zurich; Dr. Stefan M. Schmalholz, Swiss Federal Institute of Technology, Zurich; Dr. Yuri Podladchikov, University of Oslo; Dr. Erik H. Saenger,Freie Universit�t Berlin

In 2003, a group of scientists in Switzerland set out to answer some intriguing questions with implications for the way oil and gas reserves are discovered and produced. Research conducted by Dr. Stefan Dangel at the University of Zurich had highlighted a strong and consistent empirical relationship between low-frequency spectral anomalies in seismic background wavefields and geological characteristics of a collection of reservoirs, mainly in the Middle East. Similar observations have also been reported in the Russian literature since the early 1990s.

Dangel's research was robust by any standard, but focused on one feature in particular: curious amplitude peaks clustered around 3 Hz in surface velocity data measured above hydrocarbon reservoirs.¹ The possibility of a universal hydrocarbon-indicator, while attention-grabbing, did not sit well with the real-world complexities that the industry confronts day-to-day. Moreover, the reasons for such features were left largely open.

The question was whether Dangel's research pointed more generally to coherent patterns in low-frequency background waves. If so, could these be directly related to reservoirs and other subsurface structures in a way that would provide new data for exploration and production decisions?

An accumulating body of knowledge in the earth science world suggested this might be the case, but the subject had never been seriously tackled with an eye on oil and gas. The seismic industry systematically disregards seismic data below 10 Hz as noise, and for good reason: conventional geophones are insensitive in this domain and little useful data can be expected. As one geologist put it, �All my career I've been fighting noise. Now you want me to believe the noise is information?�

Yet, low-frequency waves are less susceptible to many of the problems that plague conventional seismic and electromagnetic methods, particularly in areas with poor seismic response or obstacles such as thick basalt or conglomerate layers. Successful unraveling of these patterns observed in the sub-10-Hz domain would be a valuable new contribution to exploration geophysics.

A high-quality effort would need to acquire new, high-quality datasets and tackle the physical mechanisms behind these �hydrocarbon micro-tremors� that Dangel¹ and others^2,3,4,5 have found. A strong scientific team, substantial research funding and the support of credible operating partners would also be required. All of this came at a time when investment in geophysical services, not to mention funding for technology start-ups, was at a low ebb.

With barely a scent of the present exploration boom in the air, Spectraseis Technology Inc. was founded in early 2003 to begin the task of acquiring low-frequency seismic data and to develop industrial applications as the research progressed. Promising early work with Petrobras in Brazil and a Shell affiliate in Austria drew Swiss government funding for a dedicated research group at ETH Zurich. An investment by the new technology ventures group of Norsk Hydro in 2005 helped to accelerate and expand the development of commercial acquisition systems and data processing software.

Today, with a research and development team of 10 scientists and commercial land surveys planned or in progress with Petrobras in Brazil, Pemex in Mexico, Norsk Hydro in Libya and KOC in the Arabian Peninsula, it is evident that low-frequency analysis will be part of the exploration and reservoir characterization toolkit of the future. The questions now are which applications will prove most useful and how quickly the rest of the industry will embrace them.

HYDROCARBON MICROTREMORS

The starting point for our work has been the empirical observations of Dangel, et al.,¹ and since then, augmented with larger scale surveys. More than 20 studies at different oil and gas fields around the world have shown characteristic spectral anomalies with a high degree of correlation to the location and geometry of hydrocarbon reservoirs.^1,2,3,4,5,6 

The studies focus on passive seismic data in the 1- to 10-Hz frequency range, acquired using high-sensitivity broadband seismometers instead of conventional geophones. The key observation is that modifications of the seismic background spectrum are different for interactions with geological structures containing hydrocarbon-filled pores compared to interactions with similar structures containing water only. In other words, �hydrocarbon tremors� may be viewed as a frequency dependent �scattering� of the incoming background waves. Subsurface structures, such as hydrocarbon reservoirs, generate characteristic patterns in the Fourier spectra of surface velocities, Fig. 1. 

Fig. 1. Data from a survey in Brazil showing consistent anomalies in the Fourier spectra of surface velocities, measured within and outside the boundaries of a known oil reservoir.

 

The assessment of subsurface structures by analyzing the Fourier spectra of surface velocities is in line with an increasing number of methods that investigate ambient seismic waves to get information about subsurface structures. For example, earthquake hazards for buildings are assessed in this way, by deriving site-specific amplification factors (micro-zonation) for incoming earthquake waves.

DATA ACQUISITION

Passive low-frequency surveying, as the name implies, relies on ambient seismic waves, a continuous natural wave field found with varying amplitudes at every location on earth.⁷ The main driving force for these waves are believed to be the ever-present ocean waves. They have a peak around 0.2 Hz, as can be seen in Fig. 2.

Fig. 2. A survey of seismic stations worldwide shows that seismic background waves vary in amplitude but are continuously present. The so-called �ocean wave peak� is seen around 0.2 Hz. Source: Aki, K., Richards, P.G., Quantitative Seismology: Theory and Methods. Freeman 1980.

 

As with other passive methods, low-frequency surveys enjoy significant environmental, logistical and cost advantages over conventional active seismic and electromagnetic methods that require a powerful artificial source. They also have an inherently low health-and-safety risk profile.

Spectraseis' proprietary approach, dubbed HyMAS, uses ultra-sensitive, portable 3C broadband seismometers to acquire complex data, which can be filtered to separate artificial and surface-generated signals from spectral patterns related to subsurface structures. The instruments are installed in shallow holes about 0.5-m deep for weather shielding and to improve coupling, Fig. 3.

Fig. 3. A portable broadband seismic station installed at a Pemex field in the Burgos basin, Mexico.

 

Others have experimented with different equipment. A low-frequency survey led by Shell in Saudi Arabia's Rub Al-Kali Desert employed amplified 3C, 4.5-Hz geophones, surface deployed and equipped with a low-noise amplifier to increase sensitivity and integrate passive recordings within a conventional 2D seismic acquisition workflow.⁸ 

Ideally, a grid layout is used with node spacing ranging from 250 to 1,000 m. Several monitoring stations are installed for the duration of the entire survey. The crew moves the recording stations from point to point, leaving them at a given location to measure and record between 3 and 24 hours, depending on local noise conditions (high cultural/industrial noise levels require longer measuring times), Fig. 4.

Fig. 4. A survey layout for Norsk Hydro over a field in Libya includes monitoring stations, a 1,000-m grid and two densely-spaced line profiles for the non-permanent stations.

 

Quality control is applied in the field. Data points must meet certain specifications. An efficiently managed survey of 100 sq km can be acquired by a crew of 20 to 30 people in about 30 days. This is significantly faster than the acquisition of active seismic data. Specialized data management software and field stations launched by Spectraseis in 2006 have moved low-frequency surveys from the realm of scientific experiments to a tightly controlled and highly scalable commercial operation. What does this mean for an operator? Low-frequency data can be acquired, processed, interpreted and used in decision-making in as little as 10% of the time needed to acquire and process a conventional seismic survey.

CAUSES OF 1 TO 10 HZ SPECTRAL ANOMALIES 

It has been established in many parts of the world that coherent patterns relating to oil and gas reservoirs exist in the low-frequency domain. Identifying the underlying physical mechanisms of these so-called Direct Hydrocarbon Indicators (DHIs) is the key challenge for us as researchers. Success will enable us to perform more realistic numerical simulations and sensitivity studies. Three candidates are described below and illustrated in Fig. 5.

Fig. 5. Three possible mechanisms that generate DHIs in the background spectrum: standing wave resonance, selective attenuation, resonant amplification.

 

Standing wave resonance. This occurs on a macro-scale, where characteristic maxima are generated due to reflections between the reservoir and the surface, and within the reservoir, caused by complex impedance contrast due to the reservoir. When seismic waves propagate from one medium into another medium with different complex impedance, then a part of the wave is reflected. The characteristic two-way travel time, or resonance frequency, between the Earth's surface and the bottom of a low-velocity surface layer or the top of a reservoir, generates characteristic spectral anomalies. Importantly, the effective impedance contrast can be enhanced or solely generated by high attenuation in reservoir rocks.⁹ We study spectral anomalies generated by standing wave resonance for elastic and poroelastic media.

Selective attenuation. Characteristic minima are due to frequency-dependent attenuation within the reservoir. Frequency-dependent reflections take place if the seismic waves hit a layer with diffusive attenuation properties or if waves propagate from an elastic into a poroelastic medium. ^9,10 There exist several physical models to describe the attenuation of seismic waves due to wave-induced porous fluid flow.¹¹ These models describe wave attenuation on different spatial and temporal scales.

A model that presumably describes the dominant mechanism between 1 and 10 Hz is the so-called patchy saturation model. ^12,13,14 We study patchy saturation effects within the reservoir to determine under what conditions a selective, frequency-dependent attenuation could generate spectral anomalies similar to the observed hydrocarbon microtremor signal.

Wave attenuation on the small-reservoir scale will be approximated by an effective model, capturing the essentials of wave attenuation and dispersion, at the larger scale (top 10 km) to evaluate the transfer of the spectral anomalies toward the Earth's surface. We particularly focus on the differences between gas and oil pore-fill and the consequences on the frequency dependence of the reflection coefficient.

Resonant amplification. Resonant amplification effects of ambient seismic waves are also promising candidates for explaining hydrocarbon micro-scale tremor signals. These effects will behave like a driven source and they are supported by the following observations ¹:

1.An often narrow-frequency range of the signal (1.5�4Hz)

2.The mean absolute power of the hydrocarbon tremor depends on the level of the environmental noise

3.The power of the signal is proportional to the total hydrocarbon-bearing layer thickness of the reservoir

4.3C recordings show a trough instead of a peak in the H/V-ratio

5.Investigations on the wave-field propagation directions (using a directionally sensitive sensor setup) showed that the signals causing the anomaly originate from the reservoir direction.

Direct numerical simulations using Navier-Stokes equations show that pores, which are partially saturated with oil and gas, exhibit a resonance frequency. This resonance mechanism can be approximated by a damped-oscillator model. Depending on pore geometry, the oscillator models are either linear or nonlinear. ¹⁵ Similar resonance effects have been described for capillary trapped oil blobs. ^16,17

We couple the oscillator model to a one-dimensional wave propagation equation and study under what conditions the resonance of the oil-filled pores is activated, and under what conditions the resonance frequency can be measured at the surface. In particular, we investigate the coupling between the porous reservoir material and the ambient rock material, and the subsequent propagation of the resonant waves to the Earth surface.

An alternative explanation of the mechanism is the so-called �drop-bubble model�, which considers non-equilibrium phase transitions in hydrocarbon reservoirs.⁵

DATA PROCESSING AND INTERPRETATION

The aim of processing the data is to remove or attenuate all signals that are not related to subsurface structures�mainly surface noise, generated by road traffic, industrial activities, wind and rain�and to correct the dataset for inter-temporal and near-surface geology-related variations. For this purpose, a proprietary toolbox has been developed that encompasses data storage, data management and data-selection routines. A range of techniques can be applied to the selected data encompassing time domain, frequency domain and combined temporal-spectral analysis.

The company's processing software suite, called RIO (Fig. 6), contains more than fifty processing and analysis routines, several of which are the subject of key patent applications. One recently developed, patent-pending technique applies an innovative auto-normalization method to resolve signal variations over time.

Fig. 6. RIO�s integrated workflow automates many processing functions and has drastically cut the turnaround time and strengthened quality control for low-frequency data sets.

 

Processing has been largely automated so that relatively simple cases can be processed with a minimum of human interaction. More difficult problems still need an experienced analyst to solve them. Overall, the RIO software package has drastically reduced the turn-around time for data sets and strengthened quality control by moving stable processes from the test bench to a well structured and controlled workflow.

The company has also developed a specialized mapping and geostatistical program for interpretation, which it will be providing to survey customers under a limited license, Fig. 7. It allows an efficient interpretation of low-frequency survey results. The low-frequency data can be easily and interactively overlaid with other geological and geophysical information, i.e., seismic contour maps, fault maps, etc. In addition, a variety of mathematical operators for multi-attribute cross-correlation can be applied. The program also allows a visual inspection of various data sets.

Fig. 7. A mapping tool for multi-attribute and geostatistical interpretation of low-frequency passive seismic data.

 

APPLICATIONS

Low-frequency data can be applied in exploration, field appraisal and production. It is an additional tool to reduce the risk of drilling a dry hole by providing a DHI. It should not be viewed as a stand-alone tool, as it cannot provide the detailed geometrical information necessary for planning wells.

This new passive technique is particularly efficient for exploration. It allows large exploration areas to be screened quickly and at a low cost to identify areas with a high hydrocarbon potential. More expensive 3D seismic can then be limited to these high-potential areas. This saves time and money and can significantly shorten the time span from exploration to production.

The technique is also suitable to identify stratigraphic traps that are normally not mappable on 2D- or 3D-seismic data, an increasingly important application as the oil industry runs out of classical structural traps.

Another interesting area of application is in deep offshore exploration, where large fields are still expected to be found. 3D seismic normally cannot identify whether mapped structures contain hydrocarbons. The new method can.

CASE STUDIES

The volume of data supporting low-frequency spectral analysis is now accumulating rapidly. At least nine new surveys will be completed in the first half of 2007, which will more than double the volume of data available for further research. While the usual limitations on disclosure of customer data have constrained the number of published case studies, two can be mentioned here.

Petrobras Mossoro. A 100 sq km blind test for Petrobras in 2004, covering a known complex producing field in the Potiguar basin in northeastern Brazil, clearly identified two, and partly revealed the third, producing zones within the block. The results also showed a strong positive correlation between signal amplitudes and oil column thickness measured by eight logged wells. Subsequent surface corrections and reprocessing, using newly developed 2006 processing techniques, resolved the discrepancy with respect to the third area. Further surveys starting this month in the same region will provide a comprehensive set of new results.

RAG Voitsdorf. An experimental survey in collaboration with Shell-affiliate RAG in Austria correctly predicted the company's first successful oil well in ten years. Data from an exploration target was calibrated with the results from an existing producing area some 5 km to the north. The results (Fig. 8) indicated that an oil column at least twice as thick as the 15-m payzone of the northern field could be expected from the already-planned exploration well, and in fact, some 32 m of pay was logged.

Fig. 8. The maximal value of the V/H ratio within the 1- to 6-Hz range for each sensor is shown over the southern part of a fully explored reservoir in Voitsdorf, Austria. This alternative to the standard H/V technique is an additional, proprietary attribute for microtremor hydrocarbon detection.

 

A number of major operators have also reported encouraging results from experimental surveys focusing on low frequencies, including Kuwait Oil Co. (KOC) and ADCO in the Middle East. ^2,4 

New studies from surveys in Mexico and Libya should be released in the near future, supported by both seismic and drilling results. Field development and reservoir characterization applications are the focus of a five-year technical co-operation agreement recently signed with KOC.

OUTLOOK 

2007 promises to be a breakthrough year for the development of both our scientific understanding of low-frequency behavior and interest in commercial applications of passive seismic techniques. An EAGE workshop on passive seismic methods and applications held in Dubai from December 10�14 attracted some 120 professionals from various oil companies and contractors.

Developments in the next six months will include a groundbreaking marine trial in the North Sea, in collaboration with Norsk Hydro and Scripps Institution of Oceanography in San Diego. The marine survey will use recoverable ocean bottom sensors deployed over a proven non-producing field on the Norwegian Shelf.

Our scientific strategy is to use well-established and state-of-the-art theories and tools�including data analysis, normal-mode analysis, poroelastic theory, attenuation, inversion, and numerical modeling�to evaluate and tune the technique to particular campaigns, to improve methods for finding oil and gas. Results of this research will be used to incrementally improve the data processing work flow.

Although many important questions have now been answered, others remain and will no doubt be the subject of research and debate for some years to come. The challenge for low-frequency spectral analysis, as for any new geophysical technology, will be to mature and establish the method's technical limits while channeling development efforts toward applications with the highest payoff for the industry. 

THE AUTHORS
	René Graf is chairman of Spectraseis Technology Inc. He has more than 30 years of oil and gas industry experience at Shell International and as managing partner of PROSEIS, an integrated geological and geophysical consulting firm. He holds a MSc in geophsysics from ETH Zurich. rene.graf@spectraseis.com
	Dr. Stefan M. Schmalholz is a senior research scientist and lecturer at the Geological Institute of the Swiss Federal Institute of Technology (ETH) Zurich, Switzerland. stefan.schmalholz@erdw.ethz.ch
Dr. Yuri Podladchikov is a professor and PGP Division lead at Norway’s Center of Excellence for the Physics of Geological Processes at the University of Oslo, Norway and a senior consulting scientist with Spectraseis Technology Inc. iouri.podladtchikov@fys.uio.no
Dr. Erik H. Saenger is a senior research scientist and lecturer in geophysics at Freie Universität Berlin, Germany. saenger@geophysik.fu-berlin.de