2188
Proceedings of the 18
th
International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013
4
APPLICATION OF THE EWS
The use of the EWS is summarised as follows. During a heavy
rainfall event the field data is downloaded from the site in real
time via the internet. This field data is then uploaded into the
EWS. The FOS of the previous 24 hours, using the field data, is
then obtained using the ANN. Based on the rate of change of
this predicted FOS, the time until the FOS will reach a FOS of
unity is returned to the user.
Next, the rainfall forecast for the next 5 hours for the site is
obtained from the Meteorological Service of New Zealand
(2012) via the internet. This forecast can be freely obtained by
the public. This forecast is based on the Weather Research and
Forecasting model, using data obtained from automatic weather
stations, weather radar facilities, upper air sites and marine
observation stations (Bridges, 2011). The predicted FOS over
the next 5 hours is obtained using this rainfall forecast as an
input into the ANN. The starting FOS for this predicted FOS is
the last FOS obtained using the actual sensor data. Because of
the difficulty in verifying forecasts at a local scale (Hodson,
2009), both the predicted FOS according to this forecast, and
the rate of change of the FOS obtained from the field
monitoring data, are used to estimate when failure may occur.
If failure is predicted to occur within five hours, then a stage
one warning is issued. This involves warning motorists to lower
speed limits around the landslide site. If failure is to occur
within one hour, then a stage two warning is issued. This puts a
detour in place, so motorists avoid the site altogether. Two
warnings were used because the detour route adds
approximately 25 minutes to the journey. Thus this detour route
is put in place as late as possible to avoid frustration with the
EWS due to false alarms. Warning motorists to lower speeds
around the possible landslide site in advance is intended provide
a balance between minimising the cost should the landslide
occur, and avoiding frustration at the delay to motorists. During
periods of heavy rainfall, the EWS should be updated on an
hourly basis.
5
CONCLUSIONS
A site specific EWS for rainfall induced landslides has been
developed. The EWS is based on the ability to predict the
current FOS of the site using ANNs, rainfall forecast data and
real time field measurements. The EWS proves to be useful at
predicting when failure might occur, and also returns to the user
a parameter related to the possibility of failure (the current
FOS).
A FEM was used to replicate the field response of the site to
rainfall events. This FEM was coupled with a LEA to predict
the FOS at each time-step. The results of this modelling process
were reasonably accurate, considering discrepancies caused by
natural variation within the soil and the generalised evaporation
pattern which was applied within the model.
The ANN which uses field measured data could predict the
LEA obtained FOS with good accuracy; a mean squared error of
0.41 was obtained. To predict the future FOS, an ANN using
just rainfall forecast data was developed. This ANN was less
accurate, with a mean squared error of 1.16 obtained.
It is envisioned that the methodology used to develop this
EWS can be replicated at a variety of sites as a means of risk
reduction for rainfall induced landslides.
6
ACKNOWLEDGEMENTS
The Authors would like to thank the Auckland Motorways
Alliance, Babbage Consultants, Beca Consultants and Hiway
Geotechnical for their assistance throughout the project.
7
REFERENCES
Bridges, J. (2011). MetService's Investment in Forecasting. Retrieved
10
December,
2012,
from
investment-in-forecasting/
Caine, N. (1980). The rainfall intensity-duration control of shallow
landslides and debris flows.
Geografiska Annaler Series A, 62
(1-2),
23-27.
Chae, B.-G., & Kim, M.-I. (2012). Suggestion of a method for landslide
early warning using the change in the volumetric water content
gradient due to rainfall infiltration.
Environmental Earth Sciences,
66
(7), 1973-1986.
Dai, F. C., Lee, C. F., & Ngai, Y. Y. (2002). Landslide risk assessment
and management: an overview.
Engineering Geology, 64
, 65-87.
Dhakal, A. S., & Sidle, R. C. (2004). Distributed simulations of
landslides for different rainfall conditions.
Hydrological Processes,
18
(4), 757-776.
Fredlund, D. G., Morgenstern, N. R., & Widger, R. A. (1978). The shear
strength of unsaturated soils.
Canadian Geotechnical Journal,
15
(3), 313-321.
GEO-SLOPE International Ltd. (2009a).
Seepage Modeling with
SEEP/W 2007, An Engineering Methodology, Fourth Edition
(4
ed.). Calgary, Alberta, Canada: GEO-SLOPE International Ltd.
GEO-SLOPE International Ltd. (2009b).
Stability Modeling with
SLOPE/W 2007 Version, An Engineering Methodology, Fourth
Edition
(4 ed.). Calgary, Alberta, Canada: GEO-SLOPE
International Ltd.
Harris, S. J., Orense, R. P., & Itoh, K. (2012). Back analyses of rainfall-
induced slope failure in Northland Allochthon formation.
Landslides, 9
(3), 349-356.
Hodson, A. (2009). Weather Forecast Accuracy: Part 1 Rainfall.
Retrieved
17
December,
2012,
from
ICT International Pty Ltd. (2012). Solutions for Soil, Plant and
Environmental monitoring. Retrieved 31 August, 2012, from
/
Intrieri, E., Gigli, G., Mugnai, F., Fanti, R., & Casagli, N. (2012).
Design and impementation of a landslide early warning system.
Engineering Geology, 147-148
, 124-136.
Keefer, D. K., Wilson, R. C., Mark, R. K., Brabb, E. E., Brown, W. M.,
Ellen, S. D., et al. (1987). Real-time landslide warning during
heavy rainfall.
Science, 238
(4829), 921-925.
Khanlari, G. R., Heidari, M., Momeni, A. A., & Abdilor, Y. (2012).
Prediction of shear strength parameters of soils using artificial
neural networks and multivariate regression methods.
Engineering
Geology, 131-132
, 11-18.
Lentfer, K. (2007).
Engineering geology of the Northland Allochthon,
Silverdale, North Auckland, New Zealand.
Unpublished Master
Thesis, The University Of Auckland, Auckland.
Mathworks. (2010). NARX Network (narxnet, closeloop),
Matlab Users
Guide
.
Meteorological Service of New Zealand Ltd. (2012). MetService.com -
your website for New Zealand Weather Forecasts. Retrieved 20
September, 2012, from
NIWA, & GNS Science. (2009).
Natural Hazards 2008
.
NIWA, & GNS Science. (2010).
Natural Hazards 2009
.
O'Sullivan, A. S. (2009).
Suitability of Advanced Soil Models for
Stability Analysis of Slopes in Northland Soils.
Masters Thesis, The
University of Auckland, Auckland.
Penman, H. L. (1948). Natural Evaporation from Open Water, Bare Soil
and Grass.
Proceedings of the Royal Society of London. Series A,
Mathematical and Physical Sciences, 193
(1032), 120-145.
Power, J. (2005).
Engineering Geological Properties and Slope
Movements in the Pakiri Formation and Northland Allochthon,
Mahunrangi, North Auckland, New Zealand.
The University of
Auckland, Auckland.
The MathWorks Inc. (2012). Matlab The Language of Technical
Computing. Retrieved 4 September, 2012, from
/
Van Genuchten, M. T. (1980). A closed-form equation for predicting
the hydraulic conductivity of unsaturated soils
Soil Science Society
of America Journal, 44
(5), 892-898.
