Imiriland
Impact of Large Landslides in the Mountain
 
 

WP5

B3. Workpackage description
Workpackage number 5

 

Objectives and input to workpackage

The objectives of this workpackage are:

  1. To develop hazard analysis methods based on spatial or temporal previous environmental conditions like precipitation or groundwater level, to improve the prediction of potential scenarios.
  2. To improve the neural network approach and determine its applicability to daily, weekly or monthly information, in order to predict landslide current or exceptional acceleration.
  3. To apply multivariate statistical techniques in a probabilistic approach in order to include spatial geological or geotechnical parameters in the determination of potentially catastrophic movements or in landslide zonation.

Input factors and data include:

  • Previously statistical or heuristical models developed in other contexts or which need to be improved.
  • Monitoring data collected at several sites in WP1, like daily rainfall, daily displacements, and fault distribution.
  • Climatological data in the studied areas which might be relevant to landslide behaviour.
  • Other parameters that are involved in the phenomenon like morphometric features and land use characterisation.

 

Description of work

In order to complement mechanical analyses of landslides, which may prove to be complex, the statistical and heuristical methods consist in using a large amount of previously gathered environmental data which may influence the behaviour of landslides, but in an undetermined way. In the neural network techniques, for instance, the data related to precedent rainfall, landslides displacements and velocities, in order to establish an implicit relationship which will be used for future prediction. The role of necessary filter function introduced to dampen extreme fluctuation would still need to be investigated.

In statistical techniques, not only temporal but also spatial data may be analysed, like dip, fault orientation, crack persistence, in order to determine critical criteria for landslide zoning alarm levels.

The application of such techniques, which still need to be improved and developed, will allowed a better understanding and quantification of landslide hazard, as well as provide means to fix alert criteria and acceptable levels of risk.

Such computational techniques will be used within G.I.S. environment, in order to give an added value to the processed data.

Deliverables

D12. Improved neural network code for landslide hazard assessment;

D13. developed statistical methods to produce landslide zoning and hazard quantification;

D14. determination of hazard relevant values to assess risk levels;

D15. technical manual and scientific papers;

 

Milestones and expected result

achievement of improved codes;

Testing of methods through application related to monitoring of several sites - deliverables D12 and D13 available;

Determination of hazard analysis criteria applicable to real situations - deliverable D14 available;

Integration of data within technical hazard levels - deliverable D15 available