2170
Proceedings of the 18
th
International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013
overlain by Quaternary colluvial and alluvial deposits, this
highly fractured, intensely weathered, moderately soft rock is
prone to landsliding.
In addition to the geologic setting, studies suggest that
Berkeley Hills landslide mobility is driven by precipitation and
regional active tectonic conditions (Alan Kropp and Associates
2002, Hilley et al. 2004, Quigley et al. 2010). Local orographic
precipitation forms a wet microclimate and the close proximity
to active fault traces within the San Andreas Fault zone brings
strong seismicity. Today, several hundred landslide-related
geologic and geotechnical investigation reports are available for
LBNL and the Berkeley Hills alone, and form a solid
background to this project.
3 METHODOLOGY
Two state of the art geodetic sensing technologies form the
primary modes of data acquisition in this project: high rate,
continuously streaming, GPS and InSAR. While these methods
have individually been shown capable of measuring active
ground surface displacement at scales that were previously not
possible; the appropriate characterization of landslide related
slope movement benefits from the application of both.
Where continuous GPS provides three dimensional ground
surface displacement measurements with millimeter scale
accuracy and precision at full temporal resolution, the spatial
distribution of measurement points is sparse. On the other
hand, InSAR time series analysis produces improved spatial
averages at decameter resolution with sub-centimeter precision,
and the inclusion of datasets spanning several decades of
observations. These methods are complimentary using
deformation detected across a GPS network to calibrate that
measured using InSAR.
The objective is thus to accurately measure landslide slope
deformation over time. Combining these methods allows for
spatial and temporal analysis of ground surface displacements
due to landsliding in relation to local precipitation and ground
shaking events. By incorporating these surface observations
with previous investigations and monitoring, the landslide
mechanisms can then be modeled.
3.1
GPS Data Acquisition
The first phase of this project has been to establish a network of
continuously streaming GPS stations to track landslide related
ground surface displacement over time. This involves the
instrumentation of individual landslides with autonomous,
continuously streaming GPS stations, as well as several
permanent reference stations. Each landslide station has been
specifically designed for permanent, stand-alone installation and
built to capture landslide displacement at depth. Anchored on a
deep seated reinforced concrete foundation to limit the effects
of surficial disturbance, the stations are solar powered and
equipped with a wireless antenna for remote access. Reference
station locations are chosen on the basis of proximity to the
“mobile” devices and being seated on immobile ground.
Since January 2012, 5 such “mobile” stations have been
successfully installed at LBNL and one at the University of
California Blake Garden on the Blakemont Landslide. One
reference station has been established at the Lawrence Hall of
Science above LBNL. Average daily solutions are being
obtained for each station based on a 1Hz data set, and a 25Hz
buffer is held for displacement-time histories in the case of
seismic activity. Three additional sites are in the process of
being developed.
3.2
InSAR Time Series
InSAR time series are a record of change in radar signal return
phase over time, reflecting the change in distance between the
ground surface and a satellite based radar platform (or range-
change). The strength of the return signal for each radar pulse
is dependent of the physical properties of the target (or
scatterer). Where distinct structures will return a persistent
strong signal, less prominent surfaces will return lower intensity
distributed signals and noise. Among others, two types of
InSAR time series analyses are thus known as Permanent or
Distributed Scatterer methods (respectively).
With the concurrent development of the GPS network,
analysis of InSAR time series has also begun, though is not
presented here in detail. Rather, a brief review of prior results is
described with observations based on TerraSAR-X satellite data
processed with the Tele-Rilevamento Europa (TRE)
SqueeSAR
TM
algorithm (Ferretti et al. 2011, Giannico et al.
2011).
4 PRIOR GEODETIC RESULTS
The use of InSAR time series analysis has been shown to
successfully track landslide related ground surface displacement
in the Berkeley Hills area using data sets from different
satellites over several time periods between 1992 and 2011
(Hilley et al. 2004, Quigley et al. 2010, Giannico et al. 2011).
In each case, analysis of Permanent and/or Distributed
Scatterers over the period of interest clearly exhibit accelerated
rates of displacement related to periods of high precipitation.
Though no such relationship could be established with local
seismicity, it is considered to be likely that large earthquakes
can accelerate landslide motion. Furthermore, one attempt at
the use of Continuous GPS to track landslide motion was also of
no avail (Quigley et al. 2010).
4.1
1992-2007 Time Series of ERS and RADARSAT
In the InSAR time series analysis of Permanent Scatterers
performed by Hilley et al. (2004), known landslides across the
Berkeley Hills were successfully detected and tracked from
1992 to 2001 using ERS-1 and ERS-2 data acquisitions. Over
this period, these data indicate landslide related surface
displacement rates varied between 5 to 7 mm/year range-change
in the radar line of sight direction. Based on local average slope
inclinations, this implies equivalent downslope velocities of 27
to 38 mm/year and has been verified in the field by subsurface
inclinometer displacement measurements of approximately 33
mm/year (Allan Kropp and Associates 2002). Hilley et al.
(2004) also observed that periods of landslide acceleration were
closely related to seasonal precipitation, though non-linear in
that precipitation related displacement did not occur
immediately, with lag times of up to 3 months, and did not
predictably increase with larger events. Additionally, Hilley et
al. (2004) suggest the potential for seismic related landslide
displacement given a M
w
≈ 3.9 Hayward fault event on
December 4, 1998. Though the temporal resolution of the time
series could not directly document seismically triggered
deformation, unexpectedly high InSAR displacement
measurements were observed relative to the amount of
precipitation during the same period.
Similarly, Quigley et al. (2010) examine seasonal
precipitation-related displacement, supplementing the same
ERS data set with RADARSAT-1 acquisitions from 2001 to
2006. Landslide displacement was shown not only to be of
same magnitude, but clearly seasonal and sensitive to variations
in rainfall patterns. Detrended and stacked (by month)
observations plotted against average monthly precipitation
exhibited a clear 1 to 3 month displacement response lag time
and a positive correlation to the intensity of precipitation.
4.2
2007-2009 Continuous GPS Tracking.
Quigley et al. (2010) used Continuous GPS at one known
active landslide location to track surficial displacements
between 2007 and 2009. Though InSAR time series (Hilley et
