Regional scale landslide susceptibility analysis using different gis based approaches
Sponsored Links
This presentation is the property of its rightful owner.
1 / 15

Regional scale landslide susceptibility analysis using different GIS-based approaches PowerPoint PPT Presentation


  • 139 Views
  • Uploaded on
  • Presentation posted in: General

Miloš Marjanović Department of Geoinformatics m ilos .[email protected] ý University, Olomouc. Regional scale landslide susceptibility analysis using different GIS-based approaches. Project: Methods of artificial intelligence in GIS. area. methods. materials. results.

Download Presentation

Regional scale landslide susceptibility analysis using different GIS-based approaches

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Miloš MarjanovićDepartment of Geoinformatics

[email protected] University, Olomouc

Regional scale landslide susceptibility analysis using different GIS-based approaches

Project: Methods of artificial intelligence in GIS


area

methods

materials

results

conclusion

intro

Presentation outline


intro

area

methods

materials

results

conclusion

  • Landslide Susceptibility

    likelihood of landslide occurrence over specified area or volume

  • Influence factors:

    • Triggering factors (earthquakes, rainstorms, floods etc.)

    • Natural terrain properties (lithology, relief etc.)

    • Human influence

  • Classification (Varnes, 1978)

  • Landslide mechanism (deep seated earth-slides, active & dormant)

  • Scale & detailedness


area

methods

materials

results

conclusion

intro

Fruška Gora Mountain, Serbia

  • Features (& relation to landslides)

    • Geology

    • Geomorphology

    • Hydrology

  • Landsliding history

    • 10% of the area estimated as unstable (6% dormant, 4% active, deap seated, hosted in pre-Quaternary formations)


area

methods

materials

results

conclusion

intro

  • Knowledge-driven modeling - Analytical Hierarchy Process (AHP)

  • Statistical modeling - Conditional Probability (CP)

  • Machine learning - Support Vector Machines (SVM)

  • Model evaluation measures:

    • Entropy

    • Certainty

    • Kappa-statistics

    • Area Under Curve (AUC of ROC)


area

methods

materials

results

conclusion

intro

  • Knowledge-driven modeling - Analytical Hierarchy Process (AHP)

    • Terrain attributes Xi (ranged into arbitrary class intervals)

    • Weights Wi based upon experts’ opinions

    • Addition


area

methods

materials

results

conclusion

intro

  • Statistical modeling - Conditional Probability (CP)

    • Terrain attributes Xi (ranged into arbitrary class intervals)

    • Density of landslide instances (within each class of each input terrain attribute) – Weight of Evidence

    • logit transformation and Sum


area

methods

materials

results

conclusion

intro

  • Machine learning - Support Vector Machines (SVM)

    • Classification task

    • Optimization

    • Training over sampling splits (referent data included)

    • Testing the rest of the dataset with trained classifier


area

methods

materials

results

conclusion

intro

  • Topographic maps 1:25000 (digitized to 30 m DEM)

  • Geological map 1:50000 (digitized to 30 m)

  • LANDSAT TM (bands 1-5, 2006 summer).

  • Geomorphological map 1:50000 (digitized to 30 m)

  • Arc GIS, SAGA GIS, Weka software


area

methods

materials

results

conclusion

intro

12 terrain attributes + referent landslide map:

  • Slope angle, Slope aspect, Slope length, Elevation, Slope curvature (profile and planar), Buffer of drainage network, Wetness Index

  • Lithological model, Buffer of geological boundaries, Buffer of regional structures, Referent landslide map

  • Land use map


area

methods

materials

results

conclusion

intro

AHP

CP


area

methods

materials

results

conclusion

intro

  • SVM

  • 5% of original data

  • 10% of original data

  • 15% of original data

NEW!


area

methods

materials

results

conclusion

intro


area

methods

materials

results

conclusion

intro

  • Concluding remarks and directives:

  • SVM surpassed AHP & CP by far (high performance)

  • Possible reduction of input data with similar sampling strategy

  • SVM has demanding data preparation and processing procedure

  • AHP & CP only for general insights, but GIS integrated

  • Postprocessing (smoothing out the apparent errors)

  • Preprocessing (selection of important attributes)

  • Testing on adjacent areas with incomplete data coverage


Miloš MarjanovićDepartment for Geoinformatics

[email protected] University, Olomouc

Thank you for your attention!


  • Login