Skip to content

Submarine Relief and DEM-Based Analysis

CSIR NET Earth Sciences Part C Physical Geography: continental margins, ridges, trenches, canyons, bathymetry, DEM sources, slope, D8 flow, morphometry.

  • 20 min read
  • Updated
On this page · 11 sections
  1. Prerequisites
  2. The Hook: A Map Made from Sound and Radar
  3. Submarine Relief
  4. DEM-Based Analysis
  5. Memory Tricks
  6. Don’t Confuse
  7. Common Mistakes
  8. Practice
  9. Quick Revision Table
  10. Key Takeaways
  11. What’s Next

Prerequisites

The Hook: A Map Made from Sound and Radar

Submarine Relief

The two-level Earth

Earth’s hypsometric curve is bimodal: one peak near sea level (continents, mean land elevation about 0.8 km) and one at about 4–5 km depth (ocean basins, mean ocean depth about 3.7 km). The two levels reflect two kinds of crust: thick, light, old continental crust and thin, dense, young oceanic crust floating isostatically at different heights. About 29% of the surface is land and about 71% ocean.

Provinces of the ocean floor

ProvinceDepth and gradientDescription
Continental shelfto the shelf break, typically 100–200 m deep; gradient about 0.1° (about 1:1,000)submerged continental crust; drowned river valleys, relict Pleistocene beach and dune sands; wide on passive margins, narrow on active margins
Continental slopefrom the shelf break to about 2–3 km; typically a few degrees (about 2–5°)the edge of continental crust; cut by submarine canyons; slumps and slides
Continental riseabout 2–4 km; gentleapron of sediment from turbidity currents and contour currents (contourites) at the foot of the slope; present on passive margins, absent where a trench takes its place
Abyssal plainabout 3–6 km; gradient below about 1:1,000the flattest surfaces on Earth: turbidites and pelagic sediment burying older relief
Abyssal hillsrising up to a few hundred metres above the floorfaulted volcanic crust not buried by sediment; by area the most widespread landform on the planet
Mid-ocean ridgecrest about 2.5 km deepa continuous volcanic mountain system along divergent boundaries, offset by transform faults and their inactive extensions, fracture zones
Trenchthe deepest parts of the ocean, reaching about 11 km (Challenger Deep, Mariana Trench)narrow, asymmetric troughs at subduction zones
Seamounts, guyotsrising 1 km or more from the floorisolated volcanoes; guyots are flat-topped

Active vs passive margins. A passive (Atlantic-type) margin lies within a plate: wide shelf, gentle slope, thick rise. An active (Pacific-type) margin coincides with a plate boundary: narrow shelf, steep slope, a trench instead of a rise, earthquakes and volcanoes. India’s margins are passive; the western continental shelf is markedly wider than the eastern one.

Ridge morphology and the depth–age relation

Ridge-axis relief depends on spreading rate:

SpreadingAxis formExample
Fastsmooth, broad axial high, low relief flanksEast Pacific Rise
Slowdeep axial rift valley (median valley) with rugged faulted flanksMid-Atlantic Ridge, Carlsberg Ridge
Ultraslowdiscontinuous volcanism, exposed mantle rocksSouthwest Indian Ridge, Gakkel Ridge

Away from the axis, the lithosphere cools, contracts and subsides. For ages up to roughly 70 Myr the depth follows the half-space cooling law (Parsons and Sclater, 1977):

$$\boxed{d(t) \approx 2500 + 350\sqrt{t}}$$

with \( d \) in metres and \( t \) in Myr. Older floor flattens (the plate model). This is why ridges stand high and why old ocean basins are deep.

Features built on the ocean floor

  • Seamounts are intraplate or near-ridge volcanoes. A guyot (named by H. H. Hess) is a seamount whose top was planed by waves at sea level and then carried down as the plate subsided with age.
  • Atolls follow Darwin’s subsidence sequence: fringing reef around a volcanic island → barrier reef separated by a lagoon as the island subsides → atoll once the island has sunk below the surface while the reef grows upward. Deep drilling on Pacific atolls in the mid-twentieth century reached volcanic rock beneath more than a kilometre of shallow-water limestone, confirming the model. The Lakshadweep islands are atolls on the Chagos–Laccadive Ridge.
  • Aseismic ridges are linear, seismically quiet ridges built by hotspots as plates moved over them: the Ninetyeast Ridge (Kerguelen hotspot) and the Chagos–Laccadive Ridge (Réunion hotspot) in the Indian Ocean.
  • Oceanic plateaus are thick piles of basalt from large igneous provinces (Kerguelen Plateau, Ontong Java Plateau).

Canyons, channels and fans

Submarine canyons cut the continental shelf and slope. Their heads are often opposite river mouths, and many were extended by river erosion when sea level stood about 120 m lower during glacial maxima. Their lower parts are cut and maintained by turbidity currents: dense, sediment-laden flows that run downslope under gravity. The 1929 Grand Banks earthquake set off a turbidity current whose speed was later estimated from the sequence in which it broke submarine telegraph cables.

Turbidity currents leave canyons through leveed submarine channels and build deep-sea fans. The Bengal Fan, fed by the Ganga–Brahmaputra through the Swatch of No Ground canyon, is the largest submarine fan on Earth; the Indus Fan in the Arabian Sea is another of the world’s largest.

Mapping the sea floor

MethodPrincipleNote
Single-beam echo sounderdepth \( d = v\,t/2 \) from the two-way travel time \( t \) of a sound pulse; \( v \approx 1{,}500 \) m s\(^{-1}\)sound speed varies (about 1,450–1,550 m s\(^{-1}\)) with temperature, salinity and pressure, so profiles are corrected
Multibeam (swath) echo soundera fan of beams measures a strip of depths at onceswath width \( W = 2d\tan(\theta/2) \) for a total fan angle \( \theta \); widens with depth
Side-scan sonarrecords backscatter strength along the sides of a towed fishimages texture and objects, not accurate depth
Satellite altimetrythe sea surface bulges over seamounts and sags over trenches because of their gravitygives global predicted bathymetry at kilometre-scale resolution; fills gaps between ship tracks
Airborne bathymetric LiDARgreen laser penetrates clear shallow watercoastal shallows only

Global compilations such as GEBCO combine all of these into gridded bathymetry, which is itself a DEM.

DEM-Based Analysis

What a DEM is

A digital elevation model is a digital representation of surface height. Usage of the terms varies, but in most texts:

TermMeaning
DEMgeneric: heights of a surface, usually on a regular grid
DSM (digital surface model)top of everything: canopy, buildings
DTM (digital terrain model)bare ground, with vegetation and buildings removed; sometimes includes breaklines and other terrain features

Data structures: a raster grid (one height per cell; simple, the standard for analysis), a TIN (triangulated irregular network; variable density, follows breaklines), and contours (the traditional map form, interpolated to grids).

Sources

SourceMethodNotes
SRTM (Shuttle Radar Topography Mission, February 2000)C-band radar interferometry (InSAR)about 60° N to 56° S; 1 arc-second (about 30 m) and 3 arc-second (about 90 m) grids; a surface model in forests
ASTER GDEMoptical stereo from the ASTER sensor1 arc-second grid; artefacts in cloudy areas
CartoDEMoptical stereo from ISRO’s Cartosat-1Indian national DEM, distributed through Bhuvan
ALOS World 3Doptical stereo (PRISM)about 30 m global product
TanDEM-Xbistatic X-band InSARabout 12 m and coarser global products
LiDARlaser ranging from aircraft or drones; multiple returns allow a bare-earth DTM under vegetationsub-metre resolution; local coverage
Photogrammetry and structure-from-motionoverlapping photographs from aircraft or dronescentimetre to metre resolution

Pixel size from arc-seconds. One arc-second of latitude is about \( 6{,}371{,}000 \times \pi/648{,}000 \approx 30.9 \) m everywhere; one arc-second of longitude is \( 30.9\cos\varphi \) m at latitude \( \varphi \). Geographic-coordinate DEMs therefore have rectangular, not square, cells away from the equator, and must be projected (or corrected) before slope is computed.

Accuracy is reported as the root-mean-square error of DEM heights against check points:

$$\text{RMSE} = \sqrt{\frac{1}{n}\sum (z_{\text{DEM}} - z_{\text{ref}})^2}$$

The mean error (bias) and RMSE differ: a DEM can have zero mean error and still a large RMSE. Errors grow on steep slopes, in forests (radar and optical DEMs see the canopy), in shadow, and in voids filled by interpolation.

Resolution matters. Coarser cells smooth the terrain: computed slopes become gentler, small valleys and ridges vanish, and drainage density falls. Always state the resolution when quoting a DEM-derived slope.

Primary terrain attributes

For a 3 × 3 window of cell size \( \Delta \), with the central cell \( z_5 \) and neighbours labelled

\( z_1 \) (NW)\( z_2 \) (N)\( z_3 \) (NE)
\( z_4 \) (W)\( z_5 \)\( z_6 \) (E)
\( z_7 \) (SW)\( z_8 \) (S)\( z_9 \) (SE)

the simplest (second-order finite difference) gradients are

$$\frac{\partial z}{\partial x} = \frac{z_6 - z_4}{2\Delta}, \qquad \frac{\partial z}{\partial y} = \frac{z_2 - z_8}{2\Delta}$$$$\boxed{\tan\beta = \sqrt{\left(\frac{\partial z}{\partial x}\right)^2 + \left(\frac{\partial z}{\partial y}\right)^2}}$$

(Horn’s method, used in many GIS packages, weights all eight neighbours, but the idea is the same.)

AttributeMeaningUse
Slopemagnitude of the gradient (degrees or percent; percent = 100 tan β)landslide susceptibility, erosion, soil depth
Aspectcompass direction the slope faces (direction of steepest descent)insolation, snowmelt, vegetation
Profile curvaturecurvature in the downslope directionflow acceleration (convex) or deceleration and deposition (concave)
Plan curvaturecurvature across the slopeflow divergence (spurs) or convergence (hollows)
Hillshadesimulated illumination, commonly from azimuth 315° and altitude 45°visual interpretation of lineaments and landforms
Relief, roughnessrange or standard deviation of height within a windowlandscape classification

Hydrological analysis

graph TD
    A["Raw DEM"] --> B["Fill sinks (pits)"]
    B --> C["Flow direction (D8)"]
    C --> D["Flow accumulation"]
    D --> E["Stream network: cells above a threshold area"]
    E --> F["Stream order (Strahler)"]
    C --> G["Watershed delineation from a pour point"]
    E --> H["Longitudinal profiles, slope-area, SL index, ksn"]
    G --> I["Basin morphometry, hypsometric integral"]
  1. Sink (pit) filling. Single-cell depressions, mostly artefacts, would trap flow; they are filled to the level of their lowest outlet. Real depressions (karst, glacial kettles) need care.
  2. Flow direction, D8. Each cell drains to the one of its eight neighbours with the steepest downward slope, where slope = drop/distance and the distance to a diagonal neighbour is \( \sqrt 2\,\Delta \). Multiple-flow-direction and D-infinity methods spread flow among neighbours and model divergent hillslopes better.
  3. Flow accumulation. The number of upstream cells draining through each cell; × cell area = contributing (drainage) area.
  4. Stream extraction. Cells whose contributing area exceeds a threshold are channels. A smaller threshold gives a denser network: drainage density is partly an analyst’s choice, so it should be calibrated against field or image evidence.
  5. Watersheds are traced upstream from a pour point.

Compound indices.

  • Topographic wetness index (Beven and Kirkby’s TOPMODEL): \( \text{TWI} = \ln(a/\tan\beta) \), where \( a \) is the specific catchment area (contributing area per unit contour width). High in flat hollows with large upslope areas: where saturation-excess runoff starts.
  • Stream power index: \( \text{SPI} = a\tan\beta \). High where large flows run down steep slopes: gully and channel-head sites.

Basin morphometry from a DEM

IndexFormulaInterpretation
Drainage density\( D_d = \Sigma L/A \)high on impermeable, sparsely vegetated, weak rocks
Bifurcation ratio\( R_b = N_u/N_{u+1} \)usually about 3–5; higher where structure elongates the basin
Elongation ratio (Schumm)\( R_e = (2/L)\sqrt{A/\pi} \)1 for a circle; lower for elongated basins
Circularity ratio\( R_c = 4\pi A/P^2 \)1 for a circle
Form factor (Horton)\( R_f = A/L^2 \)low for elongated basins, which have flatter, longer hydrographs
Hypsometric integral\( (\bar z - z_{\min})/(z_{\max} - z_{\min}) \)upland volume remaining; high in uplifting or resistant terrain

Tectonic geomorphology from DEMs

IndexFormulaActive tectonics indicated by
Mountain-front sinuosity\( S_{mf} = L_{mf}/L_s \) (front length along its foot / straight-line length)low values (near 1; below about 1.4): straight fronts held by active faults
Valley floor width-to-height ratio (Bull and McFadden)\( V_f = \dfrac{2V_{fw}}{(E_{ld} - E_{sc}) + (E_{rd} - E_{sc})} \)low values (below about 1): narrow V-shaped valleys incising in response to uplift
Asymmetry factor\( AF = 100\,A_r/A_t \) (\( A_r \): basin area right of the trunk stream facing downstream)values far from 50: tilting
Stream-length gradient index\( SL = (\Delta H/\Delta L)\,L \)anomalously high values: uplift or resistant rock
Normalised steepness\( k_{sn} = S\,A^{\theta_{ref}} \) with a fixed reference concavity (often 0.45)high \( k_{sn} \): fast uplift or hard rock; knickpoints where it jumps
χ (chi) analysisintegral of \( A^{-\theta} \) along the channela channel plotted against χ is straight in steady state; mismatched χ across a divide suggests divide migration

These indices are always read together with lithology: hard rock mimics uplift.

Memory Tricks

Don’t Confuse

ThisvsThat
Continental rise: sediment apron (passive margin)Trench: subduction trough (active margin)
Transform fault: active, between ridge segmentsFracture zone: inactive scar beyond the ridge offset
Guyot: flat-topped, wave-planed, then subsidedSeamount: peaked volcano never planed
Aseismic ridge (Ninetyeast): hotspot trackMid-ocean ridge: spreading axis, seismically active
Multibeam echo sounder: depthSide-scan sonar: backscatter image, not depth
DSM: canopy and buildingsDTM: bare ground
Bias (mean error)RMSE (total error magnitude)
Profile curvature: along the flow linePlan curvature: across the slope (convergence and divergence)
D8: one steepest neighbourMultiple-flow-direction / D-infinity: flow split among neighbours

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. Which margin province is typically absent where a trench lies at the foot of the slope?
  2. What is a guyot?
  3. In the D8 algorithm, to how many neighbours does a cell send flow?
Answers
  1. The continental rise.
  2. A flat-topped seamount, planed by waves at sea level and since carried below it by subsidence.
  3. One: the neighbour with the steepest downward slope.

Level 2: Part B Style

Practice Q1. The depth to ocean floor of age 36 Myr, from \( d = 2500 + 350\sqrt t \), is:

(a) 2,850 m (b) 4,600 m (c) 5,300 m (d) 15,100 m

Answer: (b) 4,600 m

\( 2500 + 350 \times 6 = 4{,}600 \) m. (a) takes \( \sqrt{t} \) as 1; (d) forgets the square root (\( 2500 + 350 \times 36 \)).

Practice Q2. A multibeam system with a total fan angle of 120° surveys a 3,000 m deep basin. The swath width is about:

(a) 3.5 km (b) 5.2 km (c) 10.4 km (d) 20.8 km

Answer: (c) 10.4 km

\( W = 2d\tan(\theta/2) = 2 \times 3000 \times \tan 60° = 6000 \times 1.732 = 10{,}392 \) m. (b) uses the half-width only; (a) uses \( \tan 30° \) (\( 2 \times 3000 \times 0.577 \)).

Practice Q3. An SRTM 1 arc-second DEM is used at 23.5° N. The approximate east–west cell size is:

(a) 30.9 m (b) 28.3 m (c) 12.3 m (d) 90 m

Answer: (b) 28.3 m

\( 30.9 \times \cos 23.5° = 30.9 \times 0.917 = 28.3 \) m. (a) is the north–south size; (d) is the 3 arc-second product.

Practice Q4. A basin outlined from a 30 m DEM drains 5,000 cells. Its contributing area is:

(a) 0.15 km\(^2\) (b) 1.5 km\(^2\) (c) 4.5 km\(^2\) (d) 45 km\(^2\)

Answer: (c) 4.5 km²

Each cell is 30 × 30 = 900 m\(^2\); \( 5000 \times 900 = 4.5\times10^6 \) m\(^2\) = 4.5 km\(^2\). (a) multiplies by 30 instead of 900.

Level 3: Part C Style

Practice Q5. A DEM is checked at eight GPS points; the errors (DEM minus GPS, m) are +1.5, −2.0, +3.0, −1.0, +0.5, +2.5, −3.5, +1.0. Which statements are correct?

  1. The mean error (bias) is +0.25 m.
  2. The RMSE is about 2.1 m.
  3. Because the bias is small, the DEM’s vertical accuracy is better than 0.5 m.
  4. If the check points lay mostly under forest, a radar DEM would be expected to show a positive bias.

(a) 1, 2 and 4 only (b) 1 and 3 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 4 only
  • 1 true: sum = +2.0; 2.0/8 = 0.25 m.
  • 2 true: squares sum to 2.25 + 4 + 9 + 1 + 0.25 + 6.25 + 12.25 + 1 = 36; \( \sqrt{36/8} = 2.12 \) m.
  • 3 false: accuracy is described by RMSE (about 2.1 m); positive and negative errors cancel in the mean.
  • 4 true: C-band radar partly reflects from the canopy, so the DEM sits above the ground (a surface-model bias).

Practice Q6. Two basins along a mountain front are analysed from a DEM:

BasinFront length along foot (km)Straight front length (km)\( V_{fw} \) (m)\( E_{ld} \) (m)\( E_{rd} \) (m)\( E_{sc} \) (m)AF
P1210501,3001,2501,00062
Q30106001,1501,1001,00049

Which statements are correct?

  1. \( S_{mf} \) is 1.2 for P and 3.0 for Q.
  2. \( V_f \) is about 0.18 for P and 4.8 for Q.
  3. P is more likely than Q to lie along an active fault.
  4. Q’s asymmetry factor indicates strong tilting.

(a) 1, 2 and 3 only (b) 1 and 4 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 3 only
  • 1 true: 12/10 = 1.2; 30/10 = 3.0.
  • 2 true: P: \( 2 \times 50/(300 + 250) = 100/550 = 0.18 \); Q: \( 2 \times 600/(150 + 100) = 1200/250 = 4.8 \).
  • 3 true: P has a straight front and a narrow V-shaped valley (both low values); Q has an embayed front and a broad valley floor, typical of an inactive front.
  • 4 false: AF = 49 is close to the neutral value of 50. P’s 62 suggests some tilting.

Practice Q7. Which statements about DEM-derived hydrology are correct?

  1. Lowering the flow-accumulation threshold increases the extracted drainage density.
  2. Sink filling is needed because pits stop flow from reaching the basin outlet.
  3. The topographic wetness index is highest on steep ridge crests.
  4. For a cell with specific catchment area 300 m and slope 5°, TWI is about 8.1.

(a) 1, 2 and 4 only (b) 1 and 3 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 4 only
  • 1 true: more cells qualify as channels.
  • 2 true: an unfilled pit is a dead end for routing.
  • 3 false: ridge crests have small contributing areas and often steep slopes; TWI is highest in flat, convergent hollows and valley floors.
  • 4 true: \( \ln(300/\tan 5°) = \ln(300/0.0875) = \ln 3429 = 8.14 \).

Quick Revision Table

IdeaFact
Hypsometrybimodal: land mean about 0.8 km, ocean mean about 3.7 km; 29% land
Shelfto 100–200 m; about 0.1°; wide on passive margins
Slope, riseslope a few degrees (about 2–5°), canyons; rise = sediment apron (passive margins)
Abyssal plain3–6 km; gradient below about 1:1,000; flattest surfaces on Earth
Ridgesfast: axial high (EPR); slow: rift valley (MAR, Carlsberg); ultraslow: SWIR, Gakkel
Depth–age\( d = 2500 + 350\sqrt t \) m (t in Myr, up to about 70 Myr)
Trenchdeepest about 11 km (Challenger Deep)
Guyot, atollwave-planed and subsided; fringing → barrier → atoll (Darwin)
Aseismic ridgesNinetyeast (Kerguelen), Chagos–Laccadive (Réunion)
Canyons and fansturbidity currents; Swatch of No Ground; Bengal Fan largest
Echo sounding\( d = vt/2 \), \( v \approx 1500 \) m s\(^{-1}\); swath \( W = 2d\tan(\theta/2) \)
DEM typesDEM generic; DSM surface; DTM bare earth; raster vs TIN
SourcesSRTM (2000, InSAR, 1″ and 3″), ASTER GDEM, CartoDEM, AW3D30, TanDEM-X, LiDAR, SfM
Arc-second1″ ≈ 30.9 m N–S; × cos φ E–W
Slope\( \tan\beta = \sqrt{(\partial z/\partial x)^2 + (\partial z/\partial y)^2} \); central differences over 2Δ
D8steepest drop ÷ distance; diagonal distance √2 Δ
Workflowfill → direction → accumulation → threshold → order → watershed
TWI, SPI\( \ln(a/\tan\beta) \); \( a\tan\beta \)
Morphometry\( D_d \), \( R_b \) 3–5, \( R_e \), \( R_c \), \( R_f \), HI
Tectonic indices\( S_{mf} \) and \( V_f \) low = active; AF far from 50 = tilt; SL and \( k_{sn} \) high = uplift or hard rock

Key Takeaways

What’s Next