← Back to project catalogue
GP-GE-1E3M4R9GeologyOpen for request

Landslide susceptibility mapping

A geospatial modelling laboratory for mapping relative landslide susceptibility from terrain, geology, rainfall, drainage, land cover, roads, and prepared landslide inventories.

  • Nuxt
  • Vue
  • FastAPI
  • PostgreSQL
  • PostGIS
  • Python
  • Rasterio
  • GeoPandas
  • scikit-learn
  • Docker

Project definition

Problem statement

Landslides result from interacting terrain, geology, water, land use, and triggering conditions. Inventory records are often clustered near roads or mapped events, which can bias a model toward surveyed locations.

The engineering problem is to prepare consistent conditioning factors, control spatial sampling and leakage, compare interpretable and learned models, and clearly separate long-term susceptibility from event-specific hazard or risk.

Project objectives

  • Prepare a documented landslide inventory with location, date, type, mapping source, and confidence where available.
  • Calculate selected slope, aspect, curvature, relief, drainage, geology, soil, rainfall, land-cover, and road-distance factors.
  • Create spatially separated training and testing samples with documented non-landslide selection.
  • Compare statistical and selected machine-learning susceptibility models.
  • Validate discrimination, calibration, spatial generalisation, and sensitivity before mapping classes.

System design

System modules

01

Inventory manager

Stores landslide geometry, type, date, source, confidence, completeness notes, and spatial split.

02

Factor processor

Creates aligned terrain, hydrological, geological, rainfall, land-cover, and infrastructure rasters.

03

Sampling engine

Generates controlled presence and non-landslide samples with spacing, exclusion, and spatial-block rules.

04

Susceptibility modeller

Runs selected logistic, tree, or ensemble methods with saved features, parameters, and calibration.

05

Map dashboard

Shows factors, inventory, susceptibility, uncertainty, validation, class areas, and optional exposure overlays.

Methodology

System workflow

  1. 01
    Define study

    The student selects a prepared hill area, mapping scale, landslide inventory, period, and validation blocks.

  2. 02
    Prepare factors

    Terrain and thematic layers are aligned, checked, transformed, and masked to the valid study area.

  3. 03
    Create samples

    Presence and comparison points are generated under documented spatial and exclusion rules.

  4. 04
    Train and validate

    Selected models train outside held-out blocks and are evaluated for discrimination and calibration.

  5. 05
    Map and test sensitivity

    The accepted model produces susceptibility and uncertainty classes under alternative input and threshold cases.

Demonstration scenario

A prepared Himalayan study area combines an inventory with terrain, geology, rainfall, land cover, drainage, and road-distance factors. Two models train outside one valley. The dashboard compares held-out performance, uncertainty, factor influence, and susceptibility classes in that valley.

Engineering

Technical architecture

Web application
Nuxt and Vue for inventory review, factor maps, model comparison, validation, and export.
Geospatial API
FastAPI for factor jobs, sampling, models, validation, sensitivity, summaries, and tiles.
Spatial data layer
PostgreSQL and PostGIS for inventories, geology, roads, drainage, samples, models, and results.
Raster pipeline
Python, Rasterio, and GeoPandas for terrain derivatives, distances, alignment, sampling, and zonal summaries.
Model pipeline
Statistical and scikit-learn methods with spatial cross-validation, calibration, importance, and reproducible seeds.

Testing

Evaluation

Evaluation measures

  • Inventory and factor spatial-quality checks
  • Area under ROC and precision-recall curves under spatial validation
  • Calibration and landslide-density agreement across susceptibility classes
  • Performance in terrain blocks excluded from training
  • Sensitivity to inventory bias, non-landslide sampling, resolution, factors, and class thresholds
  • Processing time across factor count, raster resolution, and study-area size

System boundaries

  • Susceptibility describes relative spatial likelihood under the selected historical factors and is not a time-specific landslide forecast.
  • Hazard, exposure, vulnerability, and risk require additional event, asset, and consequence information.
  • The map is not suitable for emergency warning, construction clearance, or slope certification.
  • Site decisions require detailed field mapping, geotechnical investigation, monitoring, and responsible authority review.

Included

  1. 01Landslide inventory and factor-management interface
  2. 02Terrain, sampling, modelling, validation, and sensitivity modules
  3. 03Factor, susceptibility, uncertainty, error, and exposure maps
  4. 04Prepared public geospatial data, inventories, and results
  5. 05Complete source code in a private GitHub repository
  6. 06Complete project documentation with synopsis, abstract, methodology, geological workflow diagrams, susceptibility maps, screenshots, and conclusion
  7. 07Setup and usage guide

Project record

No buyer information is collected on this page.

Permanent project ID
GP-GE-1E3M4R9
Catalogued
21 Aug 2026
Completed
Pending
Verified
Pending
Demonstration
Added when ready

Handover

After purchase

  1. 01
    Payment is confirmed

    The project is marked unavailable and cannot be purchased again.

  2. 02
    Repository access is granted

    The buyer's submitted GitHub account receives access to the private repository.

  3. 03
    The purchase record is delivered

    The certification sheet is prepared from the reviewed buyer details and sent privately by email.