Elevation Data & DEMs

A Digital Elevation Model (DEM) is simply a raster - a grid of cells - in which the value stored in each cell is an elevation in metres above a reference level, instead of a colour. Lay the grid over the land and each square answers one question: how high is the ground here. From that single surface a GIS can derive slope, aspect (which way a slope faces), hillshade (a shaded-relief picture), contour lines, drainage and viewsheds - the whole of terrain analysis, which is Module 5.

Because a DEM is a raster, everything you learned about raster resolution applies. A 30 m DEM stores one height per 30 by 30 m square; a 12.5 m DEM stores one height per 12.5 m square and so reads the terrain more finely. The cell is an average over its footprint, so a 30 m DEM cannot see a bund, a plinth or a road camber - it sees the broad shape of the land, and that is exactly what it is good for.

Two words hide inside DEM, and confusing them causes real errors. A Digital Terrain Model (DTM) is the bare earth - the ground with trees and buildings stripped away. A Digital Surface Model (DSM) is the top of everything - the height of tree canopy, rooftops and the ground where nothing stands on it.

Which you want depends on the question. For drainage, grading and flood, you want bare-earth DTM: water runs on the ground, not on rooftops. For line-of-sight, shadowing or a first cut at building heights, the DSM is what you need. Most freely available global DEMs (SRTM, ASTER) are closer to surface models over vegetated or built areas, because radar and stereo optical methods catch the top of the canopy - so over a dense city or forest a free DEM reads high. Knowing whether your DEM is bare-earth or surface is not pedantry; it decides whether your flood analysis is right.

Three global DEMs are free and cover India. SRTM (the Shuttle Radar Topography Mission, NASA) is the classic: radar-derived, about 30 m (1 arc-second), public domain, downloadable via USGS EarthExplorer and NASA Earthdata. It covers roughly 56 degrees south to 60 degrees north, which is all of India. ASTER GDEM (NASA and Japan's METI) is a stereo-optical DEM at about 30 m, near-global, free with attribution.

The finer free option is ALOS PALSAR from the Alaska Satellite Facility: L-band radar products including radiometrically terrain-corrected tiles at about 12.5 m, free with registration and attribution (the data is JAXA's, distributed by ASF). At 12.5 m it reads slope and drainage noticeably better than the 30 m pair, which is why many Indian site studies reach for it first among the free options.

India produces its own national elevation model, CartoDEM, generated by ISRO's Space Applications Centre from Cartosat-1 stereo imagery. You download it free through Bhuvan at 30 m (1 arc-second) and 90 m postings, covering the whole country.

CartoDEM is the natural first choice when you want a home-grown, nationally consistent surface for an Indian project - and it sits comfortably beside the global DEMs for cross-checking. A sensible habit is to compare two independent DEMs over your site (say CartoDEM and ALOS PALSAR): where they agree you can trust the broad terrain, and where they diverge you have found either noise or genuinely difficult ground - forest, water, steep cliffs - that deserves a closer look or, ultimately, a field or drone survey.

Free DEMs are a gift, but they have a ceiling. At 12.5 to 30 m they describe the landform - the hill, the valley, the general fall of a plot - not the site surface. You cannot design a plinth level, a retaining wall or a stormwater invert from a 30 m DEM; those need survey-grade data at sub-metre spacing, from a total-station survey, differential GPS, or a drone (the next lesson). There is also a vertical accuracy caveat: even a good 30 m DEM may be off by several metres in height, and more over forest or steep slopes.

So use free DEMs for what they are honestly good at: reading a plot's overall slope and aspect, tracing where water will flow, spotting the low ground that floods, and setting a building in its landform. When the decision turns on centimetres, escalate to survey data - and say so on the drawing.