The variability of naturally occurring magnetic field levels: 10 Hz to 8 khz

Transcription

1 GEOPHYSICS, VOL. 75, NO. NOVEMBER-DECEMBER ; P. F7 F97, FIGS., TABLE..9/.9 The variability of naturally occurring magnetic field levels: Hz to khz Ben K. Sternberg ABSTRACT The variability of naturally occurring magnetic fields in the frequency range from Hz to khz over a period of one year was studied. Contour plots for the H x, H y, and H z components and for frequencies of,,,, and Hz were produced. Average, minimum, maximum, and the standard deviations of these fields were also calculated for distinctive time intervals. In the to khz frequency range, the noise levels are typically higher at night. In the - to -Hz frequency range, the noise levels are typically higher during the day. During midto late-summer, there is frequent thunderstorm activity, known in the southwest United States as the monsoon season. The magnetic field levels are often very high during this time period. These variability ranges can be used to estimate the lowest levels of noise that may be encountered during field surveys, which is what the authors are looking for when running controlled-source electrical method surveys. These variability ranges can also be used to estimate the highest levels that may be encountered, which is what the authors are looking for when running naturalsource electrical methods surveys, such as audio frequency magnetotelluric AMT surveys. These measurements of magnetic field strength variability show that better data for controlledsource electrical measurements can be obtained using the minimum noise level measurements, as opposed to using signal integration or signal averaging with all of the data. The minimum noise level is found by using frequency bins adjacent to the signal-frequency bin. Likewise, if one is interested in measuring the naturally occurring magnetic field data, using the maximum values during each time interval makes AMT measurements possible when the natural signal level is very low, particularly in the AMT dead zone around 5 khz. INTRODUCTION During the course of our development of new electromagnetic subsurface imaging methods, we have frequently referred to the atmospheric noise literature to estimate naturally occurring noise levels. We are primarily interested in the frequency range from around Hz to approximately khz i.e., the usual frequency range for near-surface geophysical investigation. Naturally occurring magnetic fields in this frequency range are used as signals in audio frequency magnetotelluric AMT surveys Strangway et al., 97; Vozoff, 99. These same magnetic fields are viewed as noise in controlled-source audio magnetotelluric CSAMT surveys Goldstein and Strangway, 975; Zonge and Hughes, 99. These fields are also noise in conventional ground electromagnetic geophysical surveys Keller and Frischknecht, 9; Frischknecht et al., 99; Spies and Frischknecht, 99; Fitterman and Labson, 5, as well as airborne electromagnetics AEM surveys Palacky and West, 99; Fountain, 99. The primary source for noise in this frequency range is due to sferic energy from thunderstorm activity. Lightning activity in the vicinity of the recording site appears as distinctive impulse events. Worldwide thunderstorm activity provides a background noise level, even in the absence of any recorded thunderstorms in the nearby region. This lightning energy propagates around the world with relatively small attenuation in a waveguide formed by the conductive earth and the conductive ionosphere Vozoff, 99; Garcia and Jones,. The attenuation in the earth-ionosphere waveguide varies with frequency. For example, in the frequency range of 5 khz, the so-called AMT dead band, the attenuation is higher than at other frequencies. In addition to this background noise level, there are narrow-band noise signals in the spectrum due to powerline frequency usually 5 or Hz, plus the harmonics of the powerline frequency, which can be significant up to approximately khz. Some of the references that we have used in our review of AMT noise levels included the following: Labson et al. 95 show a representative noise spectrum. The level starts at approximately ft/ Hz near Hz and then gradually decreases to approxi- Manuscript received by the Editor July 9; revised manuscript received May ; published online December. University ofarizona, Laboratory foradvanced Subsurface Imaging, Tucson,Arizona. bkslasi@ .arizona.edu. Society of Exploration Geophysicists. All rights reserved. F7

2 F Sternberg mately ft/ Hz at approximately Hz. The noise level then increases from this low to approximately ft/ Hz near khz. Macnae et al. 9 discussed noise reduction techniques. We have found that Garcia and Jones is a particularly useful reference because they show the magnetic field for several time periods during May to October, 99, which provides some understanding of how much these fields can vary over time. They show data for and khz. There is also an extensive literature on extreme-low-frequency ELF; Hz and very-low-frequency VLF;tokHz radio signals and noise Watt and Maxwell, 957; Maxwell, 9; Watt, 97; Schaning and Cumme, 99; Bannister, 9. These noise measurements are typically made with a vertical whip antenna that measures the vertical E field. The horizontal H field is then calculated using the free-space impedance, E/H 77. These calculated H fields may be different from measured H fields, particularly in the frequency range 5 khz Maxwell, 97. These papers provide a general understanding of the average magnetic field levels one may expect to encounter in field measurements. But we found the need to learn more about the short-term variations that might be encountered. The literature shows us that there are large seasonal variations e.g., winter versus summer thunderstorm seasons and large daily variations e.g., nighttime versus morning and afternoon variations. In addition, we wanted to learn more about the maximum and minimum levels that are likely to be observed during typical measurement times and not just the average level. In response to this need, we established a magnetic field monitoring station in suburban Tucson, Arizona. The coordinates of the location are:.9 latitude and.9 longitude. Magnetic field levels were continuously recorded for one year, from 7 March 7 to March. The magnetic observatory remains in operation and the seasonal and daily variations recorded continue to be similar to the data shown in this paper. The location was intentionally chosen in a suburban location, rather than a remote rural location, because near-surface geophysical surveys are often carried out in these types of areas. In a later section in this paper, we will compare the response at this suburban location in the Tucson Basin with a remote rural location in another basin west of Tucson and on a mountain north of Tucson. MAGNETIC FIELD MEASUREMENT SYSTEM A key component of the magnetic field measuring system is the sensing coil. We used toroid coils Figure to sense two components of the magnetic field. These toroid coils were originally manufactured for Conoco, Inc. now ConocoPhillips by Druid Electronics, Co. for our magnetotelluric surveying instruments. We have used these coils with varying electronic interfaces and they have been very effective sensing coils for several of our projects, both in the sub-hertz and in the audio-frequency range. The coils are balanced because the pairs of coils on the top and bottom or the left and right in Figure are identical. The output of each pair of coils is fed into a Texas Instruments INA dual differential amplifier. The center tap between the coil pairs is used as the ground reference. Two components of the magnetic field are measured by each toroid. For example, the left and right pair of coils in Figure can be used to measure the vertical magnetic field and the top and bottom pair measures the east-west field component. The resonant frequency of these coils is khz. A -k resistor from each side of the coil pair to the center tap ground was used to provide a smoothly varying response curve from Hz to khz, with a maximum response at khz. The crosstalk between the two measurement axes on these coils is very low approximately 5 db because of the excellent balance between the orthogonal pairs of coils. The internal noise level of each sensor, along with the noise level of the INA amplifier, was reduced by summing the outputs from six of these toroid/amplifiers to measure the east-west and the north-south fields. Another set of six toroid/amplifiers were summed to measure the vertical field. A plot of the calculated maximum internal noise for this system is shown in Figure. This noise curve was calculated by adding together the contributions from the thermal noise of the coil resistance, the thermal noise of the damping resistor, the INA amplifier voltage noise, and the INA amplifier current noise. We verified these calculated internal noise levels by making measurements with the sensor inside of a shielded room with three layers of steel.. ft/sqrt Hz.. Figure. Toroid coil: -cm outside diameter. Each of the four coils has, turns. The core material is mu-metal tape.. Frequency (Hz) Figure. Internal noise of the magnetic field sensing coils.

3 Variability of magnetic field levels F9 The signals for the three measurement axes are conveyed via balanced, twisted pair, shielded cables to a data-acquisition system. Band-pass filters HztokHz were used to reduce out-of-band signals. A Layla, -bit analog-to-digital converter ADC from Echo Digital Audio digitized the signals. We have used this low-cost audio ADC on many of our research projects and we have found that it outperforms laboratoryadcs costing ten times as much. We wrote TIME DOMAIN X,Y,Z E E E E E+ E RX ADC Vrms E E E Seconds FFT X,Y,Z E+7 E+ E+5 E+ E+ E+ B (ft/sqrt Hz) E+ E+ Hz HISTORY X,Y,Z E+5 B (ft/sqrt Hz) E+ E+ E+ E+ E (Hours) Figure. Magnetic field signal record. Relatively quiet interval. Top record is time-domain data over -s interval. Middle record is frequency-domain data. Bottom record is a time history of frequency-domain data for khz. TIME DOMAIN X,Y,Z E+ E+ E+ E+ E+ E+ E+ E+ E Seconds FFT X,Y,Z E+7 E+ E+5 E+ E+ E+ E+ E+ Hz RX ADC Vrms B (ft/sqrt Hz) Figure. Magnetic field signal record. Note the large number of sferies and elevated frequency-domain level. our own ASIO interface and applications software for this ADC, which is available at National Instruments, LabView, software was used for all data processing. A Blackman time-domain window was used before converting the data to the frequency domain with a fast Fourier transform FFT. An uninterruptible power supply UPS was used for power backup. This system has run continuously for well over year. The Echo Digital Layla also has digital-to-analog converter DAC outputs. We used one of these DACs to drive a power amplifier and a transmitting air-core coil placed at a precise distance from the toroid coils. This was used to produce a calibration curve. This calibration was repeated several times during the -year measurement interval and small corrections were made. The data are stored in records consisting of s of data, taken every s. All of the time-domain data and the frequency-domain data for these -s records are stored in files. In addition, a summary of the data frequency-domain magnitudes at,,,, and H is stored in another file. A complete record for day is.7 GB and for year there are more than TB of data. SHORT-TERM MAGNETIC FIELD VARIATIONS Our first objective was to capture representative samples of the variation in magnetic field levels over relatively short time intervals. Figure shows an example of the audio magnetic noise level on August. This record is for the X channel north-south. The top plot is the timedomain data for a -s interval. The middle plot is the Fourier transform of the time-domain data using a Blackman Window. The spikes on this plot are the power-line frequency and its harmonics. The level shown here is typical of fairly low-level activity. We note that the expected decrease in level near khz is not visible on this plot because of the limited resolution of this real-time display. The plot can be expanded at any time during data acquisition to zoom in on individual frequencies or a small range of frequencies. The bottom plot is the complete time history for a selected frequency, in this case the -khz amplitudes for that day. Note that the time history plot shows high levels at night and then reduced levels during the day. This day/night variation is due to changes in the earth-ionosphere waveguide. The base of the ionosphere is km higher on the day side than on the night side of the earth Vozoff, 99. In the afternoon and evening in this data set, there are increased levels due to the local monsoon thunderstorm activity, which is common during July to September in the southwest United States. Figures and 5 show records taken at slightly later times. During this recording time there was

4 F9 Sternberg thunderstorm activity in southern Arizona, but there was not any sign of thunder or lightning at the recording site. The time-domain spikes are caused by individual lightning events. Figure shows a detailed comparison of the August ::5 p.m. data set and the August :: p.m. data set. We note that when there is increased sferic noise present, all nearby frequencies have elevated levels. This will be crucial when we discuss the use of recorded minimum noise level data versus long integration times or signal averaging/stacking of the data to improve signal-to-noise ratio S/N. For example, if we are transmitting a signal that we want to measure with the best possible S/N at a frequency of 95 Hz, we can measure the noise level in nearby frequency bins and determine when the 95 Hz data signal frequency also has the lowest noise level. VARIATIONS IN MAGNETIC FIELD OVER ONE YEAR E+ E+ E+ E+ E+ E+ E+ E+ E+ E Seconds FFT X,Y,Z E+7 E+ E+5 E+ E+ E+ RX ADC Vrms B (ft/sqrt Hz) E+ E+ Hz Figure 5. Magnetic field signal record. High sferic activity level. In this section, we will present selections from the full data set, recorded continuously from 7 March 7 to March. Because of space limitations in this paper, we are not able to show all of the data and at a large enough size to see the details in the plots. All of the data plots are available at the University of Arizona Laboratory for Advanced Subsurface Imaging LASI Web site Figures 7 and show contour plots for the magnetic fields with -s measured data sets, taken every s. We decided to collectsof data only every s to limit thetotal data volume to a reasonable size. The x-axis is the number of days since 7 March 7 for days in this year. Note that was a leap year. The y-axis is time of day in hours, starting at midnight hours, until the next midnight hours. The grayscale bars show the range of signal amplitudes in ft/ Hz. Each plot was produced as follows: For each day, for each frequency,,,, and Hz, and for each component H x, H y, and H z the average value over a time interval of frequency-domain signal magnitudes was calculated. Then the minimum for these s of data and the maximum of these s of data were calculated. This process was repeated for each subsequent s of data. The original data file contains a total of 7,95, rows, corresponding to the total number of measured FFTs performed at -s intervals throughout the year. There are 5 magnetic field magnitude columns corresponding to the five frequencies, multiplied by the three field components. The plot files contain,7 rows, corresponding to the 7,95, data points divided by. Because of space limitations, only the H x data are shown in this paper. The complete data set is available on the LASI Web site. All of the plots show a distinct time of season variation, with the late summer months showing much higher levels than the winter months. This variation correlates with our late summer thunderstorm season. There is also a distinct variation with time of day. This variation correlates with sunrise and sunset. We then prepared tables showing the average, maximum, and minimum values over the full one year of recording and with three different lengths of data acquisition times s, min, and min. We have broken the year up into distinctive time intervals. B (ft/sqrt Hz).E+5.E+.E+.E+ // ::5 PM Data set // :: PM Data set.e Frequency (Hz) Figure. Comparison of low- and high-level lightning sferic activity. During each day, the intervals are midnight to : a.m. approximately sunrise, : a.m. to noon, noon to : p.m. approximately sunset, and : p.m. to the next midnight. In the - to -khz frequency range, the noise levels are typically higher at night. In the - to -Hz frequency range, the noise levels are typically higher during the day. During the year, the intervals are: 7 March 7 to 5 June 7, 5 June 7 to October 7, and October 7 to March. The interval from 5 June 7 to October 7 is a time of prominent thunderstorm activity, known in- Tucson as the monsoon season. The magnetic field levels are often very high during this time period. Figure 9 shows a summary of the H x data for -s time sections of

5 a) b) Frequency - Hz, Field - Hx, window operation - Avg /7/7 // Frequency - Hz, Field - Hx, window operation - Avg /7/7 // Figure 7. Range of magnetic field levels over one year: and Hz Log amplitude (ft/sqrt Hz) Log amplitude (ft/sqrt Hz) c) d) Variability of magnetic field levels Frequency - Hz, Field - Hx, window operation - Max /7/7 // Frequency - Hz, Field - Hx, window operation - Max /7/7 // Log amplitude (ft/sqrt Hz) Log amplitude (ft/sqrt Hz) e) f) Frequency - Hz, Field - Hx, window operation - Min /7/7 // Frequency - Hz, Field - Hx, window operation - Min /7/7 // F Log amplitude (ft/sqrt Hz) Log amplitude (ft/sqrt Hz) a) Frequency - Hz, Field - Hx, window operation - Avg /7/7 // b) Frequency - Hz, Field - Hx, window operation - Avg /7/7 // Log amplitude (ft/sqrt Hz) Log amplitude (ft/sqrt Hz) c) Frequency - Hz, Field - Hx, window operation - Max /7/7 // d) Frequency - Hz, Field - Hx, window operation - Max /7/7 // Figure. Range of magnetic field levels over one year: and Hz Log amplitude (ft/sqrt Hz) Log amplitude (ft/sqrt Hz) e) f) Frequency - Hz, Field - Hx, window operation - Min /7/7 // Frequency - Hz, Field - Hx, window operation - Min /7/7 // Log amplitude (ft/sqrt Hz) Log amplitude (ft/sqrt Hz)

6 F9 Sternberg data. Figure shows a summary of the H x data for -min time sections of data. The data for H x, H y, H z, standard deviation, and average data are available on the LASI website. VERTICAL MAGNETIC FIELD SUBURBAN VERSUS RURAL LOCATION Figure shows the magnetic field spectrum on February 7 over a -h period at Hz for the x, y, and z channels observed at the primary suburban Tucson observation site. Note that the vertical z field is approximately the same size, or slightly larger, than the horizontal x north-south and y east-west fields. For a nearly uniform or layered earth, and for plane-wave AMT signals, the vertical H z component should be much smaller than the horizontal field components Vozoff, 99. At a rural field site, the Avra Valley Geophysical Test Site west of Tucson Sternberg et al., 99, we used a second set of magnetic field sensors to measure the magnetic field levels. This site is a reasonable approximation to a homogeneous layered earth. Furthermore, cultural interference e.g., power lines and pipe lines are approximately km from this site compared with m at the suburban Tucson site. We found the expected behavior at the Avra Valley Test Site; that is, the H z vertical fields were approximately one third of the level of the horizontal fields. We then conducted magnetic field surveys around the power lines and other infrastructure power, gas, and telephone lines in the vicinity of the suburban Tucson site. The fields behaved as would be expected for a long-line source carrying current at these frequencies. The vertical magnetic field strength decreased proportional to one over the distance from the power line on both sides of the power line. Note that we are not discussing -Hz and harmonic fields here. We are discussing only the naturally occurring magnetic field noise at frequencies in between the power-line harmonic frequencies. It is clear that the background noise is being induced in these long conductors. We typically find in our field surveys that although the vertical magnetic field increases as we approach geological inhomogeneities, the peak vertical field is still smaller than the horizontal fields. However, long utility lines pick up the naturally occurring magnetic field signals over very large regions and the secondary fields from these line sources may be quite large near the conductive line. Magnetic filed () /7/7-/5/7 /5/7-//7 //7-// Hz Hx : AM : AM : PM : PM : AM of day Magnetic filed () /7/7-/5/7 /5/7-//7 //7-// Hz Hx : AM : AM : PM : PM : AM of day Magnetic filed () /7/7-/5/7 /5/7-//7 //7-// Hz Hx Magnetic filed () /7/7-/5/7 /5/7-//7 //7-// Hz Hx : AM : AM : PM : PM : AM of day Figure 9. Maximum, average, and minimum values calculated over -s intervals for four time periods each day and three time periods during the year. Top plot is for Hz and bottom plot is for Hz. : AM : AM : PM : PM : AM of day Figure. Maximum, average, and minimum values calculated over -min intervals for four time periods each day and three time periods during the year. Top plot is for Hz and bottom plot is for Hz.

7 Variability of magnetic field levels F9 Large induced currents in long conductors have been well documented. Campbell and Zimmerman 9 measured low-frequency magnetic fields at various distances perpendicular to the Alaska oil pipeline. Based on these magnetic field measurements, they interpreted currents in the pipeline of tens of amperes, with surges of hundreds of amperes, which were induced in the pipeline by low-fre- +//7.E+5 X-Channel, khz Figure. Plots of the magnetic field intensity on February 7 for the H x, H y, and H z channels..e+.e+.e+.e+.e+ :: :: :: 7:: 9:: :: :: :: 9:: :: :: + //7.E+5 Y-Channel, khz.e+.e+.e+.e+.e+ :: :: :: 7:: 9:: :: :: :: 9:: :: :: + //7.E+5 Z-Channel, khz.e+.e+.e+.e+.e+ :: :: :: 7:: 9:: :: :: :: 9:: :: ::

8 F9 Sternberg quency geomagnetic variations. A current shunt in the pipeline confirmed these large currents. Boerner et al. 9 also studied induced currents in power line systems at low frequency. MAGNETIC FIELD SIGNAL STRENGTH AT OTHER LOCATIONS We have described the variations in the audio-frequency magnetic field signal strength at one primary location in southern Arizona. We will now address the question of how these signal strengths would be expected to change at other locations. We first consider the propagation of energy from a lightning stroke to a receiver site. The lightning stroke is usually modeled as a vertical electric dipole. Beyond approximately 5 km, the waves propagate as a plane wave in the earth-ionosphere waveguide, with a vertical electric field E z component and a horizontal magnetic field H component that is orthogonal to the direction of propagation Vozoff, 99. We will address only the magnetic fields in this paper. The signal attenuation along the propagation path increases linearly with frequency from approximately. db/ km at Hz to db/ km at khz, and then decreases to db/ km at khz Maxwell, 97. This magnetic field attenuation is, for all practical purposes, independent of the ground conductivity along the propagation path at frequencies of Hz to khz. At frequencies of approximately MHz and above, the signals are affected by ground conductivity. Maclean and Wu 99 and Wu and Maclean 99 describe radiowave propagation over ground. Their algorithms are applicable to the sferic signals of interest in this paper using a vertical electric dipole source and a horizontal magnetic field receiver. We have used their algorithms to calculate the effect of varying conductivity on the propagated fields and we again verified that earth conductivity variations, within our frequency range of interest Hz to khz,do not have a significant effect on the field strength of these propagated signals. In the next section, we will also show that elevation changes do not have a significant effect on the field strength of these propagated signals. Maxwell 97 shows plots of the variation of magnetic-field signal strength as a function of latitude. The maximum signal strength is near the equator, which is closest to the largest concentration of lightning activity. The signal strength drops as we move toward the north and south poles and away from the largest concentration of lightning activity near the equator. The latitude reduction varies with frequency but is approximately db at and db at latitude. These results show that the average incident sferic signal strength usually does not vary greatly from one location to another. There are daily variations, yearly variations, and a latitude variation of db from the equator to approximately latitude; however, over a year-long period, we would expect that the average incident magnetic field level would be similar at most geophysical survey sites. This is why the literature often refers to a single representative magnetic field curve versus frequency. Local geological inhomogeneities can have an effect on the magnetic field signal strengths at a particular site. McNeill and Labson 99 give an excellent description of the effect of these inhomogeneities on the sferic signals. They showed plots of the magnetic field anomalies produced by D inhomogeneities that were published privately by Madden and Vozoff 97. Although this suite of models was produced for geophysical surveys using VLF radio stations as a source, the models are directly applicable to our sferic noise study because the source configuration is the same vertical electric dipole and the models are representative of geologic inhomogeneities that we expect to encounter in our AMT and electromagnetic in-duction surveys. The frequency used for these model calculations was khz, but the models can be scaled to another frequency using well-known electromagnetic scaling equations Frischknecht, 97. Calculated magnetic fields are shown for two cases: H-polarization, in which the D geologic structure is in one direction and the source propagation path is perpendicular to this direction; and E-polarization, in which the D structure is in one direction and the source propagation path is parallel to this direction. In the H-polarization case, the magnetic field is unaffected by the geologic structure. However, in the E-polarization case, the geologic structure can produce a substantial effect on the magnetic field. For example, Figure in McNeill and Labson 99 shows the response over a thin conductive dike m in a,- m background. The horizontal magnetic field increases by a factor of. compared with the background response. Figure in McNeill and Labson 99 shows the response over a thin conductive dike m in a,- m background. The horizontal magnetic field increases by a factor of 7 compared with the background response. These results show that local geologic inhomogeneities can alter the sferic magnetic fields at a site. We show an example of an inhomogeneous site in the next section. MAGNETIC FIELD LEVELS BASIN VERSUS MOUNTAIN RECEIVER SITE The year-long magnetic field measurements that we have discussed so far were made in the Tucson Basin. The electrical resistivity of the basin alluvium at depths of tens to hundreds of meters is typically in the range of 5 m based on our electrical resistivity measurements in these basins. The basin alluvium has a relatively uniform resistivity. The elevation of this site is 75 m. To sample the variability in magnetic field level in a different geologic environment, we set up a second set of sensors in the Santa Catalina Mountains north of Tucson, Arizona, near Rose Canyon Lake on Mount Lemmon.99 latitude,.779 longitude. The elevation is m. The Santa Catalina Mountains consist of tertiary granitic rocks and older deformed and metamorphosed rocks Force, 997. The granitic rocks would be expected to have resistivities on the order of to, m. The metamorphosed rocks typically contain zones with quite low resistivity rocks, on the order of m or less. We used the algorithms in Maclean and Wu 99 and Wu and Maclean 99 to determine whether the much higher elevation of this mountain site, compared with the basin site, could significantly alter the incident magnetic fields. Avertical electric source located in the Phoenix area was used to represent the lightning stroke. The horizontal magnetic field was calculated at the mountaintop location. As a representative example, we digitized the elevation profile along a transect from the Phoenix, Arizona, area to the Catalina Mountains, a distance of km. The elevation varied from to m. We used a resistivity of 5 m over the basin part of the profile and a resistivity of m over the mountain part of the profile. We then compared the field strength at khz for this profile versus the same profile but with constant elevation. The difference in field strengths was only. db. We then changed the resistivity of the mountain section of the profile from. to. m so that it was constant over the entire profile. This led to a difference in the -khz field

9 Variability of magnetic field levels F95 strength of. db. We therefore conclude that the dominant cause of variations in signal strength between the basin and mountain sites, at frequencies of Hz to khz, is due to local geologic inhomogeneities, not propagation of the lightning energy over different conductivities or over different elevations along the path from the lightning source to the recording site. Simultaneous measurements were made for h at the Tucson Alluvial Basin site and the Santa Catalina Mountain site on July. These two locations are separated by 7 km and the Santa Catalina Mountain site is east-northeast from the Tucson Alluvial Basin site. The recorded data at both sites was dominated by large, frequent, and correlated impulsive sferic events. During these simultaneous measurements, the major thunderstorm activity in the region was concentrated km south-southeast from both sites. Thunderstorm location maps were provided courtesy of Vaisala, Inc. Because this thunderstorm activity was approximately the same distance and in the same direction for both sites, the sites should be seeing the same source signals. The signals at the Santa Catalina Mountain site were consistently larger than the Tucson Basin site. Table shows a summary of the average H x and H y signal levels in ft/ Hz and the mountain-to-basin ratios during the -h recording time. Except for Hz, the mountain site has signal levels that are consistently approximately three times larger than the basin site. INTEGRATION OF SIGNALS VERSUS SELECTION OF MINIMUM SIGNALS TO IMPROVE S/N IN CONTROLLED-SOURCE ELECTRICAL SURVEYS In this section, we compare two strategies for obtaining the optimum S/N in controlled-source electrical methods surveys; namely, signal integration or signal averaging/stacking versus selection of a minimum signal over a specified length of time. The motivation forthis comparison was shown in Figure. In this figure we noted that when there are large noise levels present due to lightning sferic activity in the region, there is a tendency for the noise level at all nearby frequencies to be elevated. For the :: data set in Figure, all of the frequency bins show noise levels on the order to ft/ Hz. For the ::5 data set, all of the frequency bins show noise levels on the order of, ft/ Hz. For example, if we were transmitting a controlled-source signal at 95 Hz, we could independently determine the ambient noise level by monitoring the noise level in nearby frequency bins. We would then use these nearby frequency data to determine which signal in the specified time interval has the lowest noise. In Figure, the top plot shows a standard data collection record for July 7 from noon : to midnight : with data records that were s long and input to an FFT. This leads to a frequency-domain record with -Hz resolution bandwidth RBW. The average noise level in this case is ft/ Hz. The middle plot in Figure shows the same data set but with the minimum noise level determined within a time interval containing s of data. This time interval is then stepped along the data every eight points. The FFT was still calculated with a -s time-domain window i.e., a -Hz RBW. The average noise level for the minimum data in these -s time intervals is 5. ft/ Hz. The bottom plot in Figure shows a data set over the same acquisition interval, but with a time-domain window that contained s of data. This time-domain window is then stepped along the time series every s of data. This means that the data have a.5-hz RBW. The average noise level for the data in these -s integration periods is 79. ft/ Hz. The reduction in noise level due to this -s integration is a factor of., which is comparable to the theoretical reduction of noise due to signal integration or signal averaging/stacking of s/ s.. For this example, the use of data integration over an -s interval leads to a noise level that is. times larger than the use of the minimum noise value in the same -s interval. To accomplish the same reduction in noise by additional integration or signal averaging/ stacking, it would take approximately.. times as long to measure each value. In certain mission-critical applications, the worst-case S/N is the parameter that determines system performance. In this situation, for the example in Figure, selection of the minimum noise measurement leads to.5 times better S/N than signal integration; that is, a reduction in integration or signal averaging/ stacking time by a factor of approximately.5. In addition to selecting the lowest noise data in a time interval, there are other approaches to controlled-source electrical method surveys, which discard, or inverse weight, only the highest noise data and then apply signal averaging Strack, 99. There are also otheramt data analysis algorithms that can be used to significantlyenhance the quality of the data e.g., Garcia and Jones, 5,. We hope that this data set, showing the variability of the naturally occurring noise, will lead to more analysis of optimum processing techniques. Table. Basin site and mountain site summary of levels. Frequency Hz Hz Hz Hz Field Component H x H y H x H y H x H y H x H y Ratio mountain to basin Mountain.E.E.E.E.E.5E.E.E site Basin site.e.7e.e 9.E 7.5E.E.5E.E

10 F9 Sternberg + 7//7 X-Channel, khz.e+ Second FFT Window, Hz RBW.E+.E+.E+.E+ :: :: :: 5:: :: :: 9:: :: :: :: :: + 7//7 X-Channel, khz.e+ Select Minimum Value Over Sec Interval.E+.E+.E+.E+ :: :: :: 5:: :: :: 9:: :: :: :: :: + Mobile 7//7 S X-Channel, khz.e+ Second FFT Window,.5Hz RBW.E+.E+.E+ Second FFT Window, Hz RBW.E+ :: :: :: 5:: :: :: 9:: :: :: :: :: Figure. Comparison of -Hz RBW record with.5-hz RBW record and minimum value selected over a -s sliding time interval. CONCLUSIONS We have studied the variability of naturally occurring magnetic field noise in the frequency range from Hz to khz over a period of year. We have broken this time period up into distinctive time intervals. During each day, the intervals are midnight to : a.m. approximately sunrise, : a.m. to noon, noon to : p.m. approximately sunset, and : p.m. to the next midnight. In the - to -khz frequency range, the noise levels are typically higher at night. In the - to -Hz frequency range, the noise levels are typically higher during the day. During the year, the intervals are 7 March 7 to 5 June 7, 5 June 7 to October 7, and October 7 to March. The interval from 5 June 7 to October 7 is a time of prominent thunderstorm activity, known in Tucson as the monsoon season. The magnetic field levels are often very high during this time period. These ranges can be used to estimate the lowest levels of noise that may be encountered, which is what we are looking for when we are running controlled-source electrical methods surveys. These ranges can also be used to estimate the highest levels that may be encountered, which is what we are looking for when we are running natural-source electrical methods surveys such as AMT surveys. We find high vertical H z fields in suburban areas, which are interpreted to be related to ubiquitous power lines in suburban areas. We find low vertical H z fields in rural areas that are in relatively homogenous earth materials, as expected. In our tests of measurements in the relatively homogenous basin-fill site, versus measurements at the geologically heterogeneous mountain site, the mountain measurements were approximately three times larger in signal strength at most frequencies. These measurements of magnetic field strength variability show that we can obtain a better S/N for controlled-source electrical measurements by using the minimum noise level measurements, as opposed to using integration or signal averaging/stacking with all of the data. Likewise, if one is interested in measuring the naturally occurring magnetic field data, the maximum values during a time interval have much higher levels than the average. In low signal strength regions e.g., near khz, the average noise lev-

11 Variability of magnetic field levels F97 el may be lower than the sensor noise level, making successfulamt surveys impossible. We find that it is still possible to obtain useable signal strengths by using the maximum value during a reasonable recording interval. ACKNOWLEDGMENTS Druid Electronics, Co. originally manufactured the toroid coils that we used in this project for Conoco, Inc. for our magnetotelluric surveying instruments. After I left Conoco now ConocoPhillips and joined the University of Arizona, Conoco donated these coils, as well as many other geophysical instruments, to LASI at the university. Conoco s early support proved to be a major contributing factor to the success of LASI at the University of Arizona. The magnetic field recording system was built by Alex Krichenko. The MatLab processing programs were programmed by Jared Jordan. I wish to thank these students for their many contributions to LASI. Ken Cummins provided the Vaisala lightning maps and has helped educate me about lightning studies. I thank the following reviewers for their very helpful comments: Steve Cotten, Marc Vallee, and Mark Everett. This investigation was funded by the Sternberg Family. REFERENCES Bannister, P. R., 9, Summary of ELF propagation validation system effective noise measurements, 97 to 97: IEEE Journal of Oceanic Engineering, OE-9, 9, doi:.9/joe.9.5. Boerner, W. M., J. B. Cole, W. R. Goddard, M. Z. Tarnawecky, L. Shafai, and D. H. Hall, 9, Impacts of solar and auroral storms on power line systems: Space Science Reviews, 5, 9 5, doi:.7/bf. Fitterman, D. V., and V. F. Labson, 5, Electromagnetic induction methods for environmental problems, in D. K. Butler, ed., Near-surface geophysics: SEG, 5. Force, E. R., 997, Geology and mineral resources of the Santa Catalina Mountains, southeastern Arizona, A cross-sectional approach: Arizona Press. Fountain, D., 99, Airborne electromagnetic systems 5 years of development: Exploration Geophysics, Papers from the International Conference onairborne Electromagnetics AEM 9, 9, nos. and,. Frischknecht, F. C., 97, Electromagnetic physical scale modeling, in M. N. Nabighian, ed., Electromagnetic methods in applied geophysics theory: SEG, 5. Frischknecht, F. C., V. F. Labson, B. R. Spies, and W. L. Anderson, 99, Profiling methods using small sources, in M. N. Nabighian, ed., Electromagnetic methods in applied geophysics applications: SEG, 5 7. Garcia, X., and A. G. Jones,, Atmospheric sources for audio-magnetotelluric AMT sounding: Geophysics, 7, 5, doi:.9/.. Garcia, X., and A. G. Jones, 5, A new methodology for the acquisition and processing of audio-magnetotelluric AMT data in the AMT dead band: Geophysics, 7, no. 5, G9 G, doi:.9/.79. Garcia, X., and A. G. Jones,, Robust processing of magnetotelluric data in the AMT dead band using the continuous wavelet transform: Geophysics, 7, no., F F, doi:.9/ Goldstein, M. A., and D. W. Strangway, 975, Audio-frequency magnetotellurics with a grounded electric dipole source: Geophysics,, 9, doi:.9/.55. Keller, G. V., and F. C. Frischknecht, 9, Electrical methods in geophysical prospecting: Pergamon Press. Labson, V. F., A. Becker, H. F. Morrison, and U. Conti, 95, Geophysical exploration with audiofrequency natural magnetic fields: Geophysics, 5, 5, doi:.9/.9. Maclean, T. S. M., and Z. Wu, 99, Radiowave propagation over ground: Chapman and Hall. Macnae, J. C., Y. Lamontagne, and G. F. West, 9, Noise processing techniques for time-domain EM systems: Geophysics, 9, 9 9, doi:.9/.79. Madden, T. R., and K. Vozoff, 97, VLF model suite nd edition, in M. N. Nabighian, ed., Electromagnetic methods in applied geophysics; SEG. Maxwell, E. L., 97, Atmospheric noise from Hz to khz, Advisory Group for Aerospace and Research Development Conference Proceedings. McNeill, J. D., and V. F. Labson, 99, Geological mapping using VLF radio fields, in M. N. Nabighian, ed., Applied geophysics applications: SEG, 5. Palacky, G. J., and G. F. West, 99, Airborne electromagnetic methods, in M. N. Nabighian, ed.,applied geophysics applications: SEG,. Schaning, B., and G. Cumme, 99, Frequency dependence of atmospheric noise intensity from to khz at low and medium latitudes: Journal of Atmospheric and Terrestrial Physics,, 5, doi:./ Spies, B. R., and F. C. Frischknecht, 99, Electromagnetic sounding, in M. N. Nabighian, ed., Electromagnetic methods in applied geophysics applications: SEG, 5 7. Sternberg, B. K., M. F. Miletto, D. J. LaBrecque, S. J. Thomas, and M. M. Poulton, 99, TheAvra Valley Ajo Road geophysical test site: geophysical surveys, geologic data, and initial development of the test site: University of Arizona Laboratory for Advanced Subsurface Imaging Report, LASI-9-. Strack, K. M., 99, Exploration with deep transient electromagnetics: Elsevier. Strangway, D. W., C. M. Swift, and R. C. Holmer, 97, The application of audio-frequency magnetotellurics AMT to mineral exploration: Geophysics,, 59 75, doi:.9/.. Vozoff, K., 99, The magnetotelluric method, in M. N. Nabighian, ed., Electromagnetic methods in applied geophysics applications: SEG, 7. Watt, A. D., 97, VLF radio engineering: Pergamon Press. Watt, A. D., and E. L. Maxwell, 957, Characteristics of atmospheric noise from to KC: Proceedings of the Institute of Radio Engineers, 5, 77 79, doi:.9/jrproc Wu, Z., and T. S. M. Maclean, 99, Radiowave propagation over ground software: Chapman and Hall. Zonge, K. L., and L. J. Hughes, 99, Controlled source audio-frequency magnetotellurics, in M. N. Nabighian, ed., Electromagnetic methods in applied geophysics applications: SEG, 7.