Kharita, A., West, M. E., & Rodgers, A. J. (2021, December). Ground motion and Structure at Fort Greely, Alaska. In AGU Fall Meeting 2021. AGU.
Figure 1: Location of four seismometers used in this study. Additionally some faults are shown. Denali Fault hosted world's largest strike-slip fault in 2002.Inset show the extent and location in a broader perspective.
Fort Greely, Alaska is located nearly 100 miles southeast of Fairbanks. This area hosts a US military base and a segment of Trans-Alaskan Pipeline System (TAPS) which transfer crude oil from the northern tip of Alaska to southern tip. Additionally, This area has seen a significant rise in infrastructure and population over past few years.
Despite its proximity to several major and minor faults, no seismological studies have been conducted in this area (probably because of lack of permanent population here). The military base is located on a sediment valley which adds to seismic hazard in the area.
My mentor in this internship (Dr. Michael West) had deployed four new permanent seismometers in the area in roughly north-south profile with average gap of two kilometers during Nov-Dec,2020 to study in-depth ground motion and shallow structure and to complement the surrounding seismic network.
Ambient Noise Characterization
The stations GREN, GRES and PS09 are located inside a military base. There is also a pump station near stations PS09 (PS09 is derived from Pump Station 09), these pump stations consists of heavy motors that operate continuously to maintain pressure of the flow in pipeline. Moreover, there are military aircrafts that takeoff and land several times during day, so we didn't expect the site to be quite. But can this data is still useful for seismic analysis? How is the data quality at each of the individual stations.
Figure 2: Weekly spectrogram of our four stations compared with the station in located in seismically quite area. Various features observed in the spectrogram are labelled thoroughly in the figure.
To answer this question, we computed PSD plots, weekly spectrograms and data quality plots, We found that although noise levels are relatively higher compared to any quite seismic site, overall noise is well within global norms(Peterson's NHNM and NHLM). Station PS09 showed a highly intense high frequency noise which we first believed to be caused due to pumping motors. Nearby stations GREN and GRES which are located very close about 2Km away, however do not showed any such high frequency noise. This means there might be some issue in the sensor of PS09 itself. Moreover, GREN and GRES consist of high sampling seismometers whereas PS09 and K24K consist of a broadband sensor. To see if other pump station seismometers located at different points along the path of TAPS, show similar high intensity high frequency waves. We observed spectrogram of all other pump stations and found this feature is consistent, thus it was concluded that the high frequency noise is caused by heavy machinery operated in pump stations and the reason we were not seeing such noise in nearby stations such as GRES and GREN maybe because the noise may have attenuated or dominated by other kind of noises.
For the purpose of ambient noise tomography, we are interested in the natural noise which predominantly occurs in micro-seismic band (1-10s). Moreover for a reliable extraction of green's function, interstation distance should be more than 3*wavelength of the largest period. However this limit maybe relaxed to 1*wavelength. Given our small interstation distances, we decided to choose 0.5-2 Hz period for extracting green's function from the cross correlation of the continuous recordings. A comprehensive quality check was performed on a cross-correlation function before doing FTAN analysis.
Dispersion curves
Inter-station Dispersion curves were computed by applying Frequency-Time analysis (FTAN) [ Ritzwoeller and Levshin, 1998] on the stacked Cross-Correlation Function (CCFs).
For the qualitative and quantitative evaluation of dispersion curves, we follow the approach described in (Bensen et al. 2007; Yang et al. 2007) for checking the seasonal repeatability with slight modification. We first computed the dispersion curves from CCF stacks of individual months (Dec, Jan, Feb, Mar, Apr), combination of two consecutive months (Dec-Jan, Jan-Feb, Feb-Mar, Mar-Apr.), three months (Dec-Jan-Feb, Jan-Feb-Mar, Feb-Mar-Apr), four months (Dec-Jan-Feb-Mar, Jan-Feb-Mar-Apr) and finally five months(Dec-Jan-Feb-Mar-Apr). Each dispersion curve was manually inspected and dispersion curves showing irregular behavior and/or spurious jumps in velocity were removed. The mean of these curves is then taken as a final dispersion curve and their standard deviation is taken to represent the uncertainties in the measurements of group velocities. These dispersion curves are supposed to represent the variation of average group speeds with depth in the area between the two stations in the pair . The smooth decrease in velocity with period might indicate the presence of unconsolidated sediments while the increase of velocities in southern direction may suggest the decrease in the thickness of sedimentary layer from north to south.
In this study, we estimated site response using three methods involving spectral ratios – (i) Standard Spectral Ratio (SSR) of S wave from earthquakes relative to a reference site, (ii) Horizontal to Vertical Spectral Ratio (HVSR) of S-wave from earthquakes and (iii) HVSR of ambient noise, also known as microtremor HVSR.
Site response measurements are a way of quantifying such frequency dependent amplifications produced as a result of combination of factors including local geology and velocity structure.
In the recent years, spectral ratios are emerging as a most popular way of measuring site response. These methods are based on the assumption that most of the ground motion at a site is caused by vertically propagating shear waves and their energy is recorded in the horizontal components. Thus, when normalized by the vertical component, horizontal components can yield the frequencies at which local amplification has occurred.
SSR, HVSR (earthquakes) and HVSR (noise) computed for all the four stations in the study were plotted (Figure5) and the peaks for all the ratios appeared to be in good agreement. All the sites showed multiple peaks. The lowest frequency peak in each case was identified and was taken as the resonant frequency of the site. The HVSR (earthquake) for a particular site had maximum amplitude among all the others and seem to be nearly double that of the SSR, this was expected as SSR indicates the amplification relative to a reference site. The SSR and HVSR (noise) appeared to have nearly same amplitude. Maximum HVSR (earthquake) amplitude of 7.5 was seen at GRES, followed by GREN and PS09 which had nearly equal amplitude of 5 and K24K had a least amplitude as expected. The predominant resonance frequencies are measured to be 0.308, 0.218, 0.168 and 0.734 Hz respectively for GREN, GRES, PS09 and K24K. If we consider the standing wave explanation of a fundamental mode resonance of a layer over halfspace, the relationship between the fundamental resonance frequency and layer thickness is expressed as -
h = Vs/4f0
Where h is the thickness of a layer, f0 is the fundamental frequency of the site and Vs is the average shear wave velocity in the layer(Kramer 1996). Using this relationship we find that the thickness of sedimentary layer is maximum beneath PS09 and is nearly four times of that of thickness beneath K24K where it is minimum. The thickness first progressively increases southward from GREN to PS09 and then decreases till K24K. Combining the elevation data with the thickness estimates we produce the sedimentary map.
Figure 6: Final SSR was computed as a SNR weighted stack. Individual SSR curves were computed from the earthquakes corresponding to different azimuths. Individual SSR was smoothed using Konno-Ohmachi filter before stacking. Dominant frequencies were selected as a frequency corresponding to maximum amplitude.
In this study we intend to provide the first ever detailed seismic analysis of the Fort Greely Area from the perspectives of seismic noise, shallow velocity structure and site response using continuous data from four recently deployed seismometers. Since the area has not been attempted for any previous high resolution broadband experiment, our research forms the groundwork for future seismic risk assessment and ground structure related studies in the area. If more stations are deployed in future, Dispersion curves from their data can be validated using our results and inverted to form the period-velocity maps which can be further used to extract the shallow 1-D shear wave velocity profile. Site response results will impact the future geotechnical investigations and will also be useful for the policy makers and disaster management authorities to create necessary risk mitigation strategies.
AGU Fall Meeting, 2021