What DEM resolution should I use for terrain analysis?

Match the cell size to the smallest landform that matters for your decision. Use 1-2 m LiDAR for site-scale slope stability and drainage, 5-10 m for catchment and corridor work, and 30 m (SRTM/Copernicus GLO-30) for regional screening.

Does higher DEM resolution always give better results?

No. Finer cells resolve vegetation, buildings, and noise that can swamp the regional terrain signal, and they multiply slope values upward. The best resolution is the coarsest one that still captures the landform you need to measure.

How does cell size change slope and drainage results?

Slope computed over a larger cell averages relief across more ground, so mean and maximum slope generally fall as resolution coarsens. Drainage networks also simplify: small channels and pits disappear at 30 m that are sharp at 1 m.

How do I resample a DEM to a different resolution?

Use gdalwarp with -tr for the target cell size and -r to choose the method: bilinear or cubic for smooth continuous elevation, and average when aggregating to a much coarser grid. Avoid nearest-neighbour for elevation.

DEM Resolution: Choosing 1 m vs 10 m vs 30 m

Short answer

DEM resolution (the ground distance one cell covers) should be chosen to match the smallest landform your decision depends on, not the largest number you can download. Use 1-2 m LiDAR-derived models for site engineering, slope stability, and detailed drainage; 5-10 m for catchments and infrastructure corridors; and 30 m (SRTM, Copernicus GLO-30, ASTER) for regional screening. Resolution is not a neutral display setting: it directly changes computed slope, the density of the drainage network, and ruggedness metrics, so a derivative is only comparable to another at the same cell size.

What resolution actually controls

Cell size sets the spatial wavelength of terrain you can represent. The Nyquist intuition applies: to capture a landform you need cells roughly half its width or smaller. A 5 m wide gully is invisible in a 30 m DEM, partly resolved at 10 m, and crisp at 1 m. Coarsening does not just blur, it integrates. Slope at a 30 m cell is the gradient across 30 m of ground, so it averages out the steep micro-relief a 1 m cell would expose. Empirically, mean and maximum slope on the same hillside fall as you move from 1 m to 10 m to 30 m, because each step folds more terrain into one number.

That has direct consequences:

Slope and aspect soften with coarser cells. A landslide-relevant 40-degree scarp on a 1 m DEM may read as 22 degrees at 30 m.
Drainage and flow accumulation simplify. Small first-order channels and closed depressions vanish; the extracted network looks tidier but loses real structure.
Ruggedness indices (TRI, roughness) scale with cell size and are meaningless to compare across resolutions.
Volumes and cut/fill depend on cell area; mixing resolutions in a difference calculation produces nonsense.

Higher resolution is not automatically more truthful

A 1 m photogrammetric or LiDAR DSM captures tree canopy, buildings, vehicles, and sensor noise. If your question is "where are the regional structural lineaments," that detail is pure interference, and a 30 m bare-earth model answers better and faster. High resolution helps only when (a) the landform you care about is small and (b) the DEM is a bare-earth DTM, not a first-return DSM full of vegetation. Always confirm whether you have a DTM or a DSM before reading slope on fine data.

Matching resolution to the decision

Decision	Typical landform scale	Suggested resolution	Common source
Slope stability at a single site	metres	0.5-2 m	LiDAR DTM, drone photogrammetry
Drainage and culvert sizing	5-50 m	1-5 m	LiDAR
Infrastructure corridor screening	10-100 m	5-10 m	national LiDAR / 10 m DEM
Catchment hydrology	50-500 m	10-30 m	Copernicus GLO-30
Regional geomorphology / first-pass screening	hundreds of m to km	30 m	SRTM, GLO-30, ASTER GDEM

Worked example: resampling and comparing

Suppose you have a 1 m LiDAR DTM and want a 10 m product for corridor analysis. Resample with an aggregating method, not nearest neighbour:

gdalwarp -tr 10 10 -r average -t_srs EPSG:32633 dtm_1m.tif dtm_10m.tif

-r average is appropriate when you are coarsening substantially, because it integrates the fine cells into each coarse cell, which is physically what lower resolution represents. For modest changes or upsampling continuous elevation, -r bilinear or -r cubic keep the surface smooth; never use -r near for elevation because it picks a single cell value and creates stair-stepping.

Now compute slope at both resolutions and compare the distributions:

gdaldem slope dtm_1m.tif slope_1m.tif -compute_edges
gdaldem slope dtm_10m.tif slope_10m.tif -compute_edges

You will see the 10 m mean slope is lower and the long tail of steep cells is compressed. That is expected, not an error, and it is exactly why you must report the resolution alongside any slope statistic. In QGIS the equivalent path is Raster > Projections > Warp (Reproject) for resampling and Raster > Analysis > Slope for the derivative; in ArcGIS Pro use Resample (Data Management) and the Slope tool.

A note on effective vs nominal resolution

The cell size in the header is the nominal resolution. The effective resolution, the finest detail actually present, can be coarser if the DEM was interpolated up from sparse points or resampled from a lower grade. A 5 m DEM upsampled from 30 m SRTM has 5 m cells but 30 m information. Check the source lineage; if the data was upsampled, do not trust slope detail finer than the original grid.

Common pitfalls and why they happen

Comparing slope or ruggedness across two resolutions. The metrics are resolution-dependent by construction, so the comparison measures cell size, not terrain. Resample both to a common grid first.
Reading slope on a 1 m DSM. Canopy and rooftops inject false steep edges. Use a bare-earth DTM for slope.
Differencing DEMs of different cell size. Cut/fill and change-detection require identical grids; resample to the coarser one before subtracting.
Assuming finer is the safe default. Finer multiplies data volume (a 1 m DEM is ~900x the cells of a 30 m one) and amplifies noise; for regional questions it is slower and noisier with no analytical gain.

QA validation

Confirm the DEM is a DTM not a DSM when slope matters; verify nominal vs effective resolution from the lineage; resample to a single common grid before any cross-DEM comparison; report the cell size next to every slope, drainage, or ruggedness figure; and sanity-check that max slope is physically plausible for the terrain at that resolution.

Bathyl perspective

The resolution question is really a scoping question: what is the smallest thing the client must be able to see or measure? Once that is fixed, the coarsest DEM that still resolves it is almost always the right choice, because it is faster, cleaner, and easier to defend. We pin the cell size to the deliverable, document the source lineage, and never let a resolution mismatch slip into a comparison.

DEM Resolution for Terrain Analysis