Short answer

For early project screening you almost always want a free, global, ~30 m DEM, and the strongest current default in most regions is Copernicus GLO-30 (TanDEM-X derived, EGM2008 vertical datum, cleaner voids than legacy data). Good alternatives to cross-check are NASADEM (reprocessed SRTM), ALOS AW3D30 (optical, often better in steep terrain), and ASTER GDEM v3 (near-global but noisier). All of these are surface models that include canopy and buildings, so they are fit for relief, drainage, and slope-class screening, not for site-grade measurement. Confirm each one's resolution, datum, and licence before you build anything on it.

What "screening" demands from a DEM

Early screening answers coarse questions: is this corridor mostly flat or mountainous, where do the major catchments drain, are there obvious steep zones or flood-prone lows, is the site worth a field campaign at all. Those questions tolerate ~30 m cells and metre-scale vertical noise. They do not tolerate unknown datum, undocumented voids, or a licence that blocks redistribution in a client report. So the selection criteria are: global/free coverage, honest void handling, a stated vertical datum, and a permissive licence. Absolute vertical accuracy of a few metres is usually fine at this stage.

The main free global sources

Copernicus GLO-30 (recommended default)

Derived from the TanDEM-X mission (X-band radar, acquisitions ~2011-2015), 1 arc-second (~30 m) globally, with a 90 m sibling (GLO-90). Vertical datum is EGM2008. Void handling and edge quality are generally better than SRTM. It is a DSM (includes surface objects). Open licence suitable for most commercial use; check the current terms. This is the sensible starting point unless you have a regional reason not to.

NASADEM

A reprocessing of the 2000 SRTM mission with improved void filling and refined elevations, 1 arc-second, covering roughly 60N to 56S. Vertical datum EGM96. If you have historical SRTM workflows, NASADEM is the modern drop-in. Distributed via NASA Earthdata (free account required).

SRTM (legacy)

The original Shuttle Radar Topography Mission, C-band radar, 30 m (SRTMGL1) and 90 m, coverage 60N to 56S. Reliable and ubiquitous but has genuine voids in steep terrain (radar layover/shadow) and is superseded for new work by NASADEM and GLO-30. EGM96 datum.

ALOS World 3D - 30 m (AW3D30)

JAXA optical stereo product, ~30 m, near-global. Often performs better than SRTM in steep mountainous terrain and is a useful independent check because its error structure differs from radar DEMs. EGM96 datum.

ASTER GDEM v3

Optical stereo from the ASTER instrument, ~30 m, the widest latitude coverage of the set (to ~83 degrees). Noisier, with artifacts in low-contrast and cloudy areas; useful at high latitudes where radar coverage stops. EGM96 datum.

For deeper or higher-latitude work, the ArcticDEM and REMA (Antarctica) strip and mosaic products offer 2 m where available, but they are screening-grade only outside their footprints.

A practical screening workflow

  1. Define the AOI and pull GLO-30 first. Download from the Copernicus Data Space or via OpenTopography, which aggregates several of these sources behind one interface.
  2. Read the header. gdalinfo -stats glo30.tif and confirm it is geographic (EPSG:4326, degrees) and note the EGM2008 datum and the NoData value.
  3. Reproject to a working projected CRS so slope is in real units: 
    gdalwarp -t_srs EPSG:32633 -r bilinear glo30.tif glo30_utm.tif
  4. Make a quick relief and slope pass: 
    gdaldem hillshade glo30_utm.tif hs.tif -az 315 -alt 35 -compute_edges
gdaldem slope glo30_utm.tif slope.tif -compute_edges
  5. Cross-check one alternative. Pull NASADEM or AW3D30 for the same AOI and difference them with the next item. Large, structured differences flag void fills or terrain where one source is unreliable. 
    gdal_calc.py -A glo30_utm.tif -B nasadem_utm.tif --calc="A-B" --outfile=diff.tif
  6. Record source, date, resolution, datum, and licence in the project metadata before any of this reaches a deliverable.

Datum: the trap that bites later

Mixing datums is the classic screening error. GNSS gives ellipsoidal heights (WGS84); global DEMs give orthometric heights (EGM96 or EGM2008 geoid). The geoid-ellipsoid separation can be tens of metres, so a "30 m error" between your field GPS and the DEM is often just the datum offset. When you later need to compare a global DEM to survey data, transform explicitly; the geoid models are well defined and the offset is computable rather than guessed.

Common pitfalls and why they happen

  • Treating a global DSM as bare earth. GLO-30, SRTM, and AW3D30 sit on canopy and rooftops; in forest the "ground" can be several metres high and slope is distorted. Fine for screening relief, wrong for site grading.
  • Ignoring SRTM/NASADEM latitude limits. Coverage stops at ~60N/56S; for high-latitude AOIs use ASTER, ArcticDEM, or REMA instead.
  • Comparing two global DEMs without matching datum and grid. Differences then conflate datum offset, resolution, and real terrain. Reproject both, confirm datums, then difference.
  • Assuming free means redistributable. Most are permissive, but confirm the licence allows including derived figures in a client report.

QA validation

Confirm the source name, acquisition era, resolution, and vertical datum are recorded; that the AOI falls inside the source's latitude coverage; that you reprojected before computing slope; that a second independent source agrees to within screening tolerance; and that the licence permits your intended use.

Bathyl perspective

For first-pass screening we default to Copernicus GLO-30, cross-check against NASADEM or AW3D30 in steep or vegetated ground, and write the datum into the metadata on day one. Screening DEMs are decision filters, not measurement instruments: they should be good enough to decide whether a site earns a survey, and honest enough that the survey team knows exactly what they are inheriting.

Related reading

Sources