How do you interpret geology from a DEM?

Generate a suite of derivatives — multidirectional hillshade, slope, curvature, and a drainage network — then read landforms as expressions of process and rock strength: scarps and lineaments for faults, breaks in slope for lithological contacts, and drainage texture for permeability and structure.

Which terrain derivatives reveal faults and lithology?

Hillshade and slope expose linear scarps and triangular facets along faults; profile and plan curvature highlight breaks of slope at contacts; and topographic position index separates ridges from valleys. Resistant beds form ridges, weak beds form strike valleys.

Why use multidirectional hillshade for structural mapping?

A single illumination azimuth hides lineaments that run parallel to the light. Multidirectional or several orthogonal hillshades reveal structures in all orientations and reduce directional bias in fault mapping.

How does drainage pattern relate to underlying geology?

Drainage patterns are diagnostic: dendritic on homogeneous rock, trellis on folded or tilted strata, rectangular on jointed or faulted bedrock, and parallel on uniform steep slopes. Extracting the network from a DEM helps infer structure and lithology.

Terrain Analysis for Geological Interpretation

Terrain as a geological proxy

Topography is the surface expression of rock strength, structure, and the processes that erode them. Read correctly, a digital elevation model is a first-pass geological map: resistant units stand as ridges, weak units are excavated into valleys, faults show up as linear scarps and offset features, and the drainage network encodes jointing and dip. Terrain analysis for geological interpretation is the disciplined extraction of these signals from a DEM using slope, curvature, hillshade, and hydrological derivatives, then cross-checking them against field and map data. This article covers which derivatives carry which geological information and how to compute them honestly.

Choosing and preparing the DEM

Resolution must match the structures of interest. Regional structure (major faults, fold belts) reads well on Copernicus GLO-30 or even GLO-90; metre-scale features (small scarps, terrace risers, landslide backscars) need lidar-derived DTMs from programmes like USGS 3DEP. Note that GLO-30 is a digital surface model: vegetation and buildings will overprint the bare-earth signal, which matters in forested terrain.

Always reproject to a metric projected CRS before computing derivatives — a UTM zone such as EPSG:32630, or a national grid. Slope, curvature, and aspect computed on a geographic CRS (EPSG:4326) are distorted because the cell is not square on the ground.

gdalwarp -t_srs EPSG:32630 -tr 30 30 -r bilinear srtm.tif dem_utm.tif

Hillshade for structure: kill the azimuth bias

Hillshade is the workhorse of structural geomorphology because the human eye reads relief from shading instantly. The trap is azimuth bias: with light from the default 315 degrees, any lineament trending NW–SE runs along the light and becomes invisible, while NE–SW features are exaggerated. Mapped fault populations come out biased toward the illumination-perpendicular orientation.

Two defences:

Generate multidirectional hillshade, which combines several light sources: gdaldem hillshade dem_utm.tif hs_multi.tif -multidirectional.
Or produce hillshades at orthogonal azimuths (e.g. 045 and 315) and inspect both before drawing any lineament.

Set -z 1 only in a metric CRS where horizontal and vertical units agree; otherwise relief is grossly exaggerated. For subtle features, a low sun altitude (-alt 25) lengthens shadows and sharpens scarps.

Slope and curvature for contacts and lithology

A lithological contact often appears as a break of slope: a resistant sandstone over a soft shale produces an abrupt steepening. Two derivatives make these breaks explicit:

Slope (gdaldem slope -alg Horn, degrees) maps where the surface steepens — useful for outlining caprock edges and dip slopes.
Curvature separates the shape of the break. Profile curvature (curvature along the slope direction) flags convex ridge crests and concave valley floors; plan curvature (across-slope) picks out spurs and hollows. In QGIS, the SAGA Slope, Aspect, Curvature tool outputs both; ArcGIS Spatial Analyst has a Curvature tool with profile and planform components.

Dipping strata create asymmetric ridges (cuestas and hogbacks): a long gentle dip slope and a short steep scarp slope. Aspect plus slope reveals this asymmetry, and the dip direction can be inferred from which face is the dip slope. Treat such inferences as hypotheses to confirm with bedding measurements.

Faults and lineaments

Map faults from a combination of evidence rather than a single layer:

Linear scarps and triangular facets in hillshade and slope.
Offset drainage and ridges — a stream that jogs sharply across a line suggests strike-slip displacement.
Aligned springs, vegetation lineaments, and breaks in topographic texture.

Quantify orientation by digitising lineaments and plotting a rose diagram; clustering of azimuths often corresponds to regional joint or fault sets. Always flag that DEM lineaments are candidates — anthropogenic features (field boundaries, tracks, pipelines) and DEM artifacts (striping, tile seams) mimic geological lines.

Drainage networks as a structural tool

Drainage pattern is one of the most reliable terrain clues to geology. Extract the network from the DEM and classify the pattern:

Dendritic — branching, tree-like; homogeneous, flat-lying or massive rock.
Trellis — parallel main streams with right-angle tributaries; folded or tilted sedimentary strata, following strike valleys.
Rectangular — right-angle bends; jointed or faulted bedrock controlled by two fracture sets.
Parallel — uniform steep slopes of consistent dip.
Radial / annular — domes, volcanoes, eroded folds.

A standard extraction in QGIS or GRASS:

Fill sinks (GRASS r.fill.dir or QGIS Fill sinks (Wang & Liu)).
Flow direction and flow accumulation (r.watershed, or QGIS Catchment area).
Threshold accumulation to define channels and vectorise (GRASS r.to.vect or Channel network).

Drainage density (channel length per unit area) is itself diagnostic: high density indicates impermeable, easily eroded lithologies (clays, shales); low density indicates permeable or resistant rock (sandstone, limestone karst).

Worked example: reading a folded sedimentary terrain

Reproject GLO-30 to UTM, generate multidirectional hillshade and slope in degrees.
Identify long parallel ridges with consistent asymmetry — cuestas dipping in one direction; the scarp faces give dip direction.
Extract drainage; observe a trellis pattern with strike-aligned subsequent streams in the soft beds.
Note a rectangular kink in one valley crossing the ridges — candidate fault; confirm with offset of the ridge crest.
Overlay the published geological map: ridges should coincide with mapped resistant formations. Where they do not, you have either a mapping error to investigate or a DEM artifact to discount.

Common pitfalls and why they happen

Trusting a single hillshade — azimuth bias systematically rotates your mapped fault population. Use multidirectional.
DSM versus DTM confusion — interpreting canopy edges as geological scarps in forested GLO-30 data.
Tile seams and striping read as lineaments — always cross-check suspicious straight lines against the DEM's processing history.
Over-interpreting low-resolution data — a 90 m DEM cannot resolve a 30 m fault scarp; the apparent feature may be interpolation.
Ignoring vertical exaggeration — a wrong z-factor invents relief that drives false interpretation.

QA and validation

Reconcile interpreted contacts and faults against the published geological map and any borehole or field data.
Check lineament populations for anthropogenic and artifact contamination.
Re-derive key layers at a second resolution; features that vanish at coarser scale should be flagged as scale-dependent.

Bathyl perspective

We use terrain derivatives as a reconnaissance layer that focuses fieldwork, not a substitute for it. Each interpreted structure carries the evidence that supports it — scarp, drainage offset, slope break — so a reviewer can weigh it. Topography proposes; mapping and measurement confirm.