Landslide Mapping with Mismatched Data
A district of 1,200 km² straddles two UTM zones. Available data include a 30 m DEM, 1:50,000 topographic sheets, a 1:250,000 geological map, 1:10,000 cadastral parcels in a local datum, a 5 km rainfall grid and a landslide inventory of 214 points for 2015-2024. The inventory has problems: 17 points fall on roads, 9 fall in rivers and 6 are duplicates. The final map is required at 1:25,000 for village planning. In a trial overlay, landslide points sit up to 180 m from the mapped slope break. The team proposes to resample everything to 30 m and publish without field validation because only six field days remain.
1. Diagnose the likely causes of the 180 m offset and the inventory errors. Explain which causes are georeferencing problems and which are attribute or classification problems.
Read the worked answer
A 180 m offset is far too large to be explained by a 30 m DEM. The most likely cause is a datum or projection mismatch: the cadastral data are in a local datum, while the landslide inventory and DEM are in UTM. A missing transformation can shift features by hundreds of metres.
The 17 points on roads and 9 in rivers are probably attribute or georeferencing errors, not real landslides. They may come from GPS readings taken without differential correction, from points digitised at a small scale, or from a classification that recorded rockfall and bank collapse as landslides. The 6 duplicates indicate no unique site ID and no checking for repeated visits.
A further cause is scale mismatch. A 1:250,000 geological map generalises boundaries by hundreds of metres; overlaying it at 1:25,000 creates false precision even when the coordinates are correct.
I would test three control points in each dataset, compare against a known benchmark, and reject the inventory points that cannot be relocated within 25 m. Only then can the slope-break offset be interpreted.
2. Recommend a projection, a mapping scale and a six-day validation design. Justify each choice in terms of the decisions village planners will make.
Read the worked answer
Projection: use one UTM zone for the whole district, with a documented transformation for the cadastral data to that zone, and record EPSG codes. If the district cannot fit one zone without unacceptable distortion, use a local transverse Mercator and say so on the map. Village planners need coordinates that match their cadastral maps, so the transformation must be explicit and tested.
Scale: publish at 1:25,000, but derive the map from 30 m DEM and 1:50,000 topography, not from the 1:10,000 cadastral detail. State that the map is for village screening, not site design. This avoids implying 1:10,000 precision that the geology and rainfall grid cannot support.
Six-day validation:
- Day 1-2: check 20 geodetic control points and correct the datum transformation.
- Day 3-4: field-check 25 inventory points, including all 17 road and river points, and delete or reclassify them.
- Day 5-6: visit 15 high-susceptibility sites near villages, recording slope, material and evidence of movement.
Use a GPS with stated accuracy and a stratified sample by susceptibility class, not a convenience sample.
3. Explain how you would communicate uncertainty on the final 1:25,000 susceptibility map without undermining its use.
Read the worked answer
Separate the map into two layers: susceptibility class and confidence. Susceptibility can be shown in five classes, but confidence should be a mask based on inventory density, geological scale and field validation. A village planner can then see that a high-susceptibility polygon in the north is supported by 12 checked landslides, while a similar polygon in the south rests on a 1:250,000 boundary and no field check.
Use fuzzy class boundaries rather than hard lines: a transition zone of 50-100 m is more honest than a single line at this scale. Put the 180 m offset problem in the metadata, state that it was corrected, and give the residual positional error after correction.
The legend should say 'screening tool: not for foundation design'. That does not undermine use; it directs planners to the right decision, such as commissioning a site survey before permitting a school. A short confidence note in the village planning pack is more useful than a single overall accuracy figure.