DEVINE -  Landslide Risk

---

La Clapiere Landslide

Abstract

Introduction

The volume of approximately 50 million cubic meters has been in motion for at least a century, but the first quantitative study in 1948 using correlated aerial images

GPS Measurement Campaign

 Three GPS receivers were installed equally spaced along the base of the landslide within the limits of the obviously mobile blocks, CLP3 being the farthest north, CLP2 in the center, and CLP1 the farthest south. Site CLP1 was located in a slide block with no vegetation. CLP2 and CLP3 were located in clearings in a series of deformed terraces with surrounding low vegetation. The choice of sites was limited by accessibility. The satellite mask caused by the slope reached up to 50 degrees for almost the entire north-eastern quadrant, but was an unavoidable characteristic of the topography. CLP1 was also located near laser target 24. Site CLP2 was located 16.7 meters from laser target 23 and 7.7 meters from a radar reflector.

The four Ashtech dual frequency Z-12 receivers recorded continuously at a sample interval of 15 seconds from 7th to 12th of April, 1999 (doy 097-102). Each site had Ashtech choke ring antennas and the reference site was also equipped with a conical radome. The antennas were mounted on wooden tripods and optically centered over the metal pins embedded in concrete which served as geodetic markers. Cumulative snowfall of 50 cm during the experiment had a small but detectable effect on the measurements.

GPS Data Processing Strategy

ECEF ITRF94 WGS-84
X 4544006.843 ? 44o 14’38.19575’’
Y 552760.550 ? 6o 56’ 08.62430’’
Z 4428424.477 H 1253.843

Table 1 Coordinates of CLPR reference site

The positions of the mobile stations (CLP1, CLP2, and CLP3 located on the landslide) were calculated relative to the reference site CLPR also using GAMIT. We used precise IGS orbits. CLPR was constrained to the above coordinates ?1 cm. Because of the short baseline lengths (700 meters maximum) we did not solve for tropospheric parameters. We modeled antenna phase center variations (Mader, ref$$). We used L1 frequency observations only (f1=1575.42 MHz) since the ionosphere-free dual frequency combination is actually noisier on short baselines (Genrich and Bock, 1992).  Dual frequency observables were used, however, for cycle slip editing and to check the resolution of ambiguities.

The session length was varied with the objective being to determine the optimal duration for confidently detecting motion of the mobile stations. We calculated baselines between the stations for sessions of decreasing length of 24 hours, 12 hours, 6 hours, and 1 hour. 24 hour long sessions were clearly long enough to average out multi-path noise and  varations due to the geometry of the satellite constellation represented by GDOP (Geometric Dilution of Precision). The sessions of 12, 6, and 1 hours demonstrate the increasing influence of biases with the decrease in measurement time.

Displacement Rates

Baseline Day 098 099 100 101 102
 Shortening by 12 mm Extension by 2 mm
CLPR-CLP1
 Shortening of 11 mm No change
CLPR-CLP2
 No change Shortening by 6 mm
CLPR-CLP3

Table 2 Change in baseline length for the three baselines over 5 days

Over the  5 day period, all of the baselines have shortened, which is the general trend expected from the geometry with the mobile sites approaching the reference site in the horizontal plane as the landslide moves down the slope (Figure 2). The size of the signal is well above the estimated short term accuracy of the measurements of 2 mm.  The evolution of the positions of the three mobile sites shows that the three sites are moving independently, with the minimum and maximum occurring at different times, as well as the simultaneous extension of the CLPR-CLP1 baseline with shortening of the CLPR-CLP3 baseline.

The evolution in time of the vertical component of the baseline length for the three mobile stations is shown in Figure 3. An increase in the vertical component of the baseline length implies a decrease in the altitude of the mobile sites since they are located at an altitude lower than the reference site. All three sites show an elevation increase followed by an elevation decrease. We do not have an independent estimate of the vertical repeatability of the measurements, but based on similar work ($$REF$$) we can expect an accuracy of approximately a centimeter. The variations in altitude of station CLP1 and CLP2 are twice this error level. The altitude of CLP1 initially increases by 2.5 cm then  descends by 3.5 cm. Station CLP2 initially increases in altitude by 1.5 cm then descends by 1 cm. CLP2 remains at approximately the same altitude with the variations not exceeding the expected error level.

A map view of the evolution in time of the three positions is shown in Figure 4. The movement of CLP1 is first towards the west and then towards the south. Based on the topography of the slope, this would predict first that the site descends and then remains at the same altitude, whereas the vertical measurements show that the site moves up then down abruptly. The movement of CLP2 is predicted by the topography to be first downslope and then remaining at the same altitude, whereas the motion is first up then followed by a descent. This apparent inconsistency may be explained by soil expansion, as depicted in Figure 5. Under this hypothesis, the largest motion is vertical motion upwards as the soil expands and then vertical motion downwards as the groundwater profile reaches the slip interface for the block and the whole block descends gravitationally with both these signals producing horizontal motion towards the facing slope. $$??!!whattaya think of that?.

Time series of temperature from the village of St.Etienne de Tinée and precipitation (snow)  and mean temperature from the ski station at Auron on the summits just above St. Etienne were collected from the Meteofrance archives. They are shown in figure Figure 9. Heavy precipitation on the order of 50 mm of snowfall preceded the experiment. The two following days (days 98-100) positive temperatures led to melting of the snow. Lower temperatures and continued snowfall put an end to the melting and the soil stabilized hydrologically $$$ is this right$$$ until the final day of the experiment.

There is some suggestion of diurnal periodicity in the time series of both the horizontal and vertical components of motion in the time series where 12 hour solution sessions are used. This may also be an artifact of the sampling in 12 hour sessions. However, if it is significant it may provide additional support for the argument of melting snow providing the triggering mechanism for slip. The daily rise in temperature and afternoon snow melt may cause an increase in the soil expansion corresponding to the upwards motion during the second session of 12 hours in Figure 7.

GPS Data Processing Optimization

We have calculated time series for the baseline length and its horizontal components for sessions of 12 hours, 6 hours, and 1 hour for comparison with the 24 hour session solution. Each 24 hour session was broken into 2 consecutive sessions of 12 hours with no overlap and no sliding window. Similarly for the 6 and 1 hour sessions, each day was broken into 4 sessions of 6 hours and 24 sessions of 1 hour respectively. These results are shown in Figure 7, Figure 8, and Figure 9.

The calculations for sessions of 12 and 6 hours show the same tendencies as the 24 hour session. The 24 hour solution appears to be a smooth version of the 6 or 12 hour series. However, as expected, the shorter the session length, the greater the noise that is superimposed on the landslide signal. For sessions less than 6 hours the discrimination of the true signal, i.e. the signal visible in the 24 hour session time series, is impossible. In addition the amplitude of the deviation from the true signal increases with shorter sessions. For sessions of 1 hour it is clear that the sign and amplitude of this noise is consistent for all three sites. These observations are also true for the vertical component of baseline length.

We investigate the correlation between multipath and the noise in the position estimates.

Discussion

Figure 1 La Clapière landslide, St. Etienne de Tinée, France. The aerial photograph taken in 1995 shows freshly exposed white rock surrounding the sliding mass on the northeast side of the valley. The locations of the GPS stations are shown as black triangles. Station CLPR on the southwest side of the valley is the stable reference station.
Figure 2 Variation in meters of the baseline length for each 24 hour session relative to initial value. Hour 0 corresponds to day 098 00:00 GMT. The time series are spline interpolated in between the 5 individual point measurements which are plotted at the middle of each session.
Figure 3 Variations in meters of the vertical components of the three baselines for 24 hour sessions. Since the mobile sites CLP1, CLP2, and CLP3 are all located at a lower elevation than the reference site, an increase in the vertical component of the baseline length implies a decrease in the altitude of the mobile site.

Figure 4 Evolution of the position of points CLP1, CLP2, and CLP3 as a function of time.

Figure 5 Hypothesis of soil expansion to explain apparent increase in elevation of station CLP1 and CLP2.

Figure 6 Variation in meters of the baseline length for each 12 hour session relative to initial value. Hour 0 corresponds to day 098 00:00 GMT. The time series are spline interpolated between successive 12 hour samples.

Figure 7 Variation in meters of the vertical component baseline length for each 12 hour session relative to initial value. Hour 0 corresponds to day 098 00:00 GMT. The time series are spline interpolated between successive 12 hour samples.

Figure 7 Comparison of the baseline length variations retrieved using several different session lengths for station CLP1. The $$ line corresponds to a session length of 24 hours, $$ corresponds to a session length of 12 hours, $$ corresponds to a session length of 6 hours, $$ corresponds to a session length of 1 hour. For session lengths of 6 hours and less, the uncertainty in the baseline length estimates obscures the trend.

Figure 8 Comparison of the baseline length variations retrieved using several different session lengths for station CLP2.

Figure 9 Comparison of the baseline length variations retrieved using several different session lengths for station CLP3.

Figure 10 Meteorological conditions during the experiment. Anomalous snowfall during the measurement period accumulated on the GPS antennas and caused an additional source of error. Melting snow on the $$last$$ day of the experiment may have caused soil expansion as water soaked into the permeable surface sediments.