Terrain Characterization using SRTM Data - Springer Link

Photonirvachak

J. Indian Soc. Remote Sens. (March 2010) 38:11–24

RESEARCH ARTICLE

Terrain Characterization using SRTM Data P. P. Patel . A. Sarkar

Received: 25.03.2009 / Accepted: 01.08.2009

Keywords SARTM . Terrain

Abstract Earth’s surface possesses relief because the geomorphic processes operate at different rates, and geologic structure plays a dominant role in the evolution of landforms (Thornbury, 1954). The spatial pattern of relief yields the topographic mosaic of a terrain and is normally extracted from the topographical maps which are available at various scales. As cartographic abstractions are scale dependent,

P. P. Patel ( ) . A. Sarkar Department of Geography, Presidency College, Kolkata -700073, India

topographical maps are rarely good inputs for terrain analysis. Currently, the shuttle radar topography mission (SRTM) provides one of the most complete, highest resolution digital elevation model (DEM) of the Earth. It is an ideal data-set for precise terrain analysis and topographic characterization in terms of the nature of altimetric distribution, relief aspects, patterns of lineaments and surface slope, topographic profiles and their visualisation, correlation between geology and topography, hypsometric attributes and finally, the hierarchy of terrain sub-units. The present paper extracts the above geomorphic features and terrain character of part of the Chotonagpur plateau and the Dulung River basin therein using SRTM data.

Introduction

e-mail: [email protected]

Survey of India topographical maps, at a variety of map-scales are the most readily available data source for terrain analysis. But, the scale-induced

12

cartographic constraints often restrict the accuracy of the topographic database (tBase) as depicted by an almost infinite variety of the patterns of contour layouts. The DEM data, on the other hand provides a very enticing and feasible option for generating a reliable database of the earth’s surface. A DEM has been defined as a regular gridded matrix representation of the continuous variation of relief over space (Burrough, 1986). The attraction of a simple matrix of elevation values was one of the most important reasons for its uptake in the early 1970s as a model suitable for landscape analysis (Evans, 1972). As mathematical–statistical models can be easily built by well-defined algorithms, digital manipulation and visualization of tBase becomes easier. Thus, as a model of surface form, the steady and widespread use of DEM may be attributed to its easy integration within a GIS environment (Weibel and Heller, 1991). The SRTM was a joint venture of National Aeronautics and Space Administration (NASA’s) Jet Propulsion Laboratory (JPL), National Imaging & Mapping Agency (NIMA), and the German and Italian Space Agencies (Werner, 2001). Using the spaceborne imaging radar (SIR–C) and X–band synthetic aperture radar (X–SAR) hardware that flew twice on the Space Shuttle ‘Endeavour’, the mission collected 12 terabytes of data covering the entire globe (600N – 600S) in February 2000 in about 10 days, mapping some of the least accessible regions of the world. The DEMs currently distributed by the United States Geological Survey (USGS) were derived from interferometric analysis of the C– band signal and were processed by NASA (Farr and Kobrick, 2000). The SRTM data are distributed at two levels % SRTM1 (for the United States and its territories / possessions) with data sampled at 1 arc-sec intervals, and SRTM3 (for rest of the world) sampled at 3 arcsec. The 3 arc-sec data were generated by (3x3) averaging of the 1 arc-sec samples and has an accuracy of ±16m, far greater than that of even a 1:25,000 scale topographical map. The SRTM3 data are divided into (10x10) tiles and distributed as height files with *.hgt extension. File names refer to the latitude and longitude of the lower left corner of the tile, e.g., N22E086 has its origin at the lower left corner

J. Indian Soc. Remote Sens. (March 2010) 38: 11–24

at (860E, 220N). Each pixel in SRTM3 data is about 90m in extent. Heights are referenced to the WGS84 / EGM96 geoid in meters and data voids are assigned a value of -32768. Each SRTM3 file contains 1201 lines and 1201 samples per line (Rodriguez, et al., 2005). The objectives of the present study were to characterize the terrain of the Dulung river basin using SRTM DEM in terms of the ¯ • nature of the altimetric frequency distribution, • relief aspects through the delineation of ‘pits’ and ‘peaks’. • patterns of lineaments and surface slope. • geometric properties of the topographic profiles and their visualisation, • correlation between underlying geology and surface topography, • hypsometric attributes of the terrain and • hierarchy of the terrain sub- units. Materials and methods Study Area The present study area lies between 22°N – 23°N and 86°E – 87°E and comprises parts of Purbi Singhbhum and Paschimi Singhbhum districts (Jharkhand), Bankura, Medinipur and Puruliya districts (West Bengal) and Mayurbhanj district (Orissa), covering the Singhbhum Shear Zone and its surroundings and drained by the Subarnarekha, Kadkal and Kangsabati rivers and their tributaries. It is a diverse landscape, with residual hills, dissected plateaus, escarpments and rolling plains, as can be deciphered from visual interpretation of the SRTM DEM of the area (Fig. 1), and thus, well suited for the ensuing analysis.

Methodology The methodology followed for the present study is as follows (Fig. 2): • Digitization of Survey of India (SOI) topographical Maps (73J – R.F. 1:250,000 for terrain characterization and 73J/10, 11, 14, 15, 16


Fig. 1 Location of the study area.

Fig. 2

Work process

13

14

• •

•

• •


– R.F. 1:50,000 for terrain correlation with geology). Procuring the SRTM-DEM file from Unites States Geological Survey (USGS) – file N22E086. Importing the STRM DEM *.hgt format files into the TAS (Terrain Analysis System) software, followed by the operations of ‘pit removal’ and ‘depression filling’ to eliminate gaps in the data. Extraction of altimetric histogram and slope profiles from the SRTM DEM to characterize the various terrain attributes of the study area and their hypsometric analysis. Contour generation and preparation of relief and slope maps of the study area from the SRTM DEM. Delineation of distinctive terrain units based on above analysis for the study area and co-relation of the same with the underlying geology obtained from Geological Survey of India, Jamshedpur Quadrangle Map 73J – R.F. - 1:250,000. for the Dulung River Basin (situated in south-eastern sector of study area).

Data pre-processing While the data coverage of SRTM is global, some regions are missing data due to lack of contrast in the radar image, presence of water or excessive atmospheric interference. Such data voids are usually along rivers, in lakes and in steep regions and often on hillsides having similar aspect due to shadowing. Any such gaps or data voids must be filled in to produce a continuous and contiguous surface before further manipulations. Often, DEMs contain extensive flats (areas of equal elevation). Usually, cells within a flat region do not have down-slope neighbours, and therefore, computation of surface slope becomes difficult (Bamler, 1999). Such ‘digital flats’ sometimes reflect minor features in the real landscape and are normally preserved during processing. If their removal allows a more robust surface generation, then they are erased after confirmatory checks using topographical maps of the area concerned. A corrected DEM was thus finally generated from the raw data for ensuing analysis.

Cell altimetric distribution The SRTM tile of the study area comprises 1,442,401 grids with elevation ranging between 10 m and 930 m. Each cell has an area of 8100 m2. Using the DEM cell values, an altimetric frequency histogram (Fig. 3) was generated with 10 m class interval. It gives the nature of the ‘area–altitude’ distribution of the study area, as follows : • It is a positively skewed frequency distribution of elevation with tail towards the higher values. • It is a multimodal distribution with principal mode in the elevation classes of (80 – 90) meters. • The auxiliary modes are located in the elevation ranges of (130 – 140 m), (150 – 160 m), (240 – 250 m), (330 – 340 m) and (400 – 410 m). • Within the topographic amplitude of 920 m, about 25% of the area lies between the elevation ranges of (30 – 110 m), 50% of the area between (30 – 170 m), and 75% of the area between (30 – 260 m). • Obviously, only 25% of the area lies between the elevation range of (260 – 930 m). Relief aspects and pit–peak delineation In any landscape, the two extremes of topographic character are the depressions or pits and the peaks or unobstructed rises in the topography, with the remaining surface grading from one to the other. Pits provide clues to flow accumulation and being the antithesis of peaks, point towards regions where these may exist. Pits are observed primarily along broad

Fig. 3 Altitudinal frequency histogram

15


stream courses and along the base slopes of prominent ridges where flowing water accumulates due to decreasing channel gradients and subsequent checks in flow velocity. These then, also indicate zones of pediment formation and possible talus fan or cone development. A relief image of the study area (Fig. 4) clearly depicts the valleys of the major streams, especially the Subarnarekha river. Its is narrow, incised, of straight configuration and steep gradient in the central part of the image, grading to broad and meandering waves in the south-eastern part which has much gentler surface gradient. The Dulung and the Kasai rivers, both occupy similar gentle valleys. The northwestern and south-western sectors appear more rugged relative to their eastern counterparts, displaying prominent ridges. Peaks are grids of high elevation that rise above the surrounding cells. The fundamental parameters for the delineation of the possible topographic peak zones are – absolute relief, threshold slope angle and

slope tolerance (Wood, 1996). The thresholds have been defined as: minimum altitude of 600 m, a gradient of 450 or more and at least 1.5 times steeper than neighbouring cells. The above criteria removes any ambiguity during peak classification that might have seen steep river banks classified as ‘peaks’ if only slope angle was considered or whole plateau surfaces being classified likewise if only absolute elevation was considered. The slope tolerance value further prunes the search for a progressively sharply rising land unit, from which a maximum viewshed can be obtained. Expectedly, peak zones fall in the rugged western half of the study area, away from the broad flat plains of the east. A large contiguous peak zone is seen in the south western sector, representing a chain of high hills while the Dalma hills show up clearly as peaks in the north (Fig. 4). Lineaments and Surface Slope The Subarnarekha valley shows up in stark contrast (in dark tint) to the surrounding areas; numerous

Fig. 4 Relief image showing major streams and terrain features

16


streams that originate from the hills and plateaus and dissect them have formed fine mesh/finger-like patterns and finally flow into the main river, especially along its left bank. Being characterized by steeper slopes, higher hills and plateaus, the south western sector of the image appears contrastingly distinctive from the south eastern sector, which comprises a vast expanse of very gentle plains, imperceptibly grading into one another. The landforms around the Kasai valley in the north-east show a mix of low dissected plateaus and small hills along with gentle and broad plains. The alternating ridge-valley configuration of the Singhbhum Shear Zone (SSZ) showed up prominently in the north and was distinct from the surrounding hills dissected lower plateau surfaces and plains. The surface slope was derived by computing the gradient of each cell. Peak areas and ridges in the west have the highest values and the eastern plains lowest values, while the intervening plateaus, the values in between (Fig. 4). Topographic profile analysis and 3D visualisation surface profile analysis done through the enumeration of the elevation values of the cells across which a profile line passes. Six ‘representative profiles’ were taken to identify the various topographic levels, terrain components and their individual characteristics (Fig. 5). The major observations (Table 1) are :

•

The Subarnarekha valley has an incised trough upstream as it cuts through the plateau floor and a low, very gentle and broad valley floor in its lower reaches. • The similar broad flat plains of the Kasai river in the north eastern sector. • The Dalma range having high altitude and steep slopes lies in the north and similar hills are also found in the southwest. • The jagged sharply rising hills and plunging incised valleys are found in the SSZ with alternating ridge–and–valley configuration. • The dissected higher plateau surfaces are characterized by steep escarpments. • The low residual hills flank the higher plateau surfaces. • The much dissected lower plateau surface gently grades into the flood plains. • The small residual hills / mounds are found in the south eastern sector. To compare the topographic levels and their progressive changes along latitudinal and longitudinal directions, sets of serial profiles were extracted and superimposed. Each horizontal or vertical profile line passes over 1201 cells covering a total distance of 108.09 km. The superimposed horizontal profiles show distinct terrain units. From west to east, these are: dissected higher plateau, higher hills, highest peaks, lower hills, dissected lower plateau and gentle plains.

Table 1 Accordant summit levels ascertained from profile analysis Topographic Unit

Elevation Range (m)

Attribute

Gentle rolling plains

below 100

Erosional plain

Lowest accordant summit level

100 – 120

Erosional plain

Second accordant summit level

140 – 160

Erosional plain

Third accordant summit level

200 – 220

Erosional plain

Fourth accordant summit level

360 – 380

Plateau rim

Fifth accordant summit level

480 – 500

Plateau rim

Sixth accordant summit level

680 – 700

Plateau proper

Seventh accordant summit level

860 – 880

Plateau proper

Highest peaks

above 880

Peak


Fig. 5

17

Profile taken across the study area

Thus, the western section is more rugged with high hills and peaks. A central zone of much dissected and incised hills lies betwixt the higher hills and plateaus of the west and rolling plains of the east. The superimposed vertical profiles show similar terrain units and reveals how the land grades between the various topographic levels. There is a stark difference

in altitude between the two adjacent accordant summit levels/planation surfaces in the southern zone of the image while the northern planation surface stands betwixt the former two, to which they both either grade up or grade down. The greater ruggedness of the southern section is emphasised while its northern counterpart has some low hills and

18

the odd high peak. The lower plains in the central area abut to either of the above. Taken together, the attributes of the horizontal and vertical profiles suggest that the south western sector is the most rugged while the south eastern sector is the most gentle. Lying adjacent to each other, they result in an abrupt rise in surface elevation and slope as one travels from the east to the west. Profile statistics have been enumerated to mathematically deduce the nature of relief along them by computing the mean elevation, range, standard deviation and grid-to grid altitudinal difference of the 1201 cells that lie along each profile line (Fig. 6). Clear distinctions are present between the computed statistics for the latitudinal profile lines (LPL) and those for the meridional profile lines (MPL). The graph of profile line-wise mean elevation reveals an almost uniform mean elevation apart from a slight dip in the middle section (over valleys) for the LPL, since these profiles traverse over both plateaus and plains. On the other hand, the mean elevation graph for the MPL reveals grading of the landscape from west to east with markedly lower mean elevations in the east where the MPL traverse almost exclusively over low plains. As a result, the corresponding range of elevations is much lower for MPL (except where some of them pass over an alternating ridge-valley zone like the SSZ), while they are much higher for LPL since they represent both topographic highs and lows. There are thus, fewer abrupt changes within the maximum and minimum cellto-cell elevation-difference lines for the MPL as their amplitude of relief is lower than that of the LPL. For the MPL, in the east the maximum and minimum elevation lines almost converge suggesting a flat landscape with very low relative relief. Standard deviation values are much lower for the MPL since they pass over much more uniform individual terrain units than across diverse terrain facets as do the LPL. Thus there is a stronger correlation between the range and standard deviation values for the MPL than for the LPL. It appears that in the study area the MPLs rather than the LPLs yield better performance in discriminating and delineating the terrain units from


west to east (following MPL) as this minimises intra terrain-unit anomalies in relief and morphometric aspects. Using DEM data like SRTM certainly makes this otherwise laborious process, quicker and simpler. Three-dimensional (3-D) visualisations generated for various sites in the study area depict the above outlined and profiled terrain units and corroborates the pertaining discussion (Fig. 7). Thus, the analysis of profiles suggests that the terrain comprise three major divisions ¯ plateau proper, plateau rim and erosional plain (Table 2). Erosional plains were further subdivided into four sub–units, plateau rim into two sub–units and plateau proper into three sub–units. Spectacular accordant summit levels are distinctive in the plateau proper and plateau rim regions. These are suggestive of uplifted peneplains / planation surfaces. Correlation between Geology and Relief Apart from regional climate, it is the regional geological structure that most controls the operations of the exogenous processes, developing thereby the existing matrix of relief. The geology of the Dulung basin, situated in the south-eastern sector of the study area has been studied and correlated with its topographic mosaic. The basin lies on the eastern fringe of the Chotonagpur plateau, just to the southeast of the SSZ. The area mainly comprises the Cenozioc sediments with some basic and ultrabasic volcanics and associated sediments formed as intrusives within the geosynclinal to platform metasediments of schists, phyllites and quartzites. The basin exposes a wide area of Tertiary and Quaternary sediments, horizontally lain with shales, conglomerates and Dhalbhum laterites. The width and thickness of the Quaternary sediments increase towards south and east of the basin. The northwest portion of the basin is geologically much more complex consisting mainly of garnetiferous phyllites of the Singhbhum Group (lower Proterozioc) that dips towards NNE at an angle of 70° while its foliation planes dip towards SE at an angle of 55°. The northwestern part of this bed is overlain by carbon – phyllite intrusives containing quartzite bands, both being the members of the Dalma Volcanics (lower Proterozioc). The carbon – phyllite beds dip towards SE at an angle of 60° while its

Fig. 6 (a) Graphs for hotizontally taken profiles (b) Graphs for vertically taken orifukes

J. Indian Soc. Remote Sens. (March 2010) 38: 11–24 19

20


Fig. 7 3-D views of portions of the study area generated from the SRTM DEM clearly depicts the different terrain faces & various planation surfaces.

foliation planes dip towards NW at an angle of 60°. There is an arm like intrusion southwestwards of garnet – staurolite schist with kyanite belonging to the Singhbhum Group. In the central part, some bands of quartzites belonging to the Dalma Volcanics are present (Fig. 8). The upstanding portion of the basin area coincides with areas of crystalline and metamorphic rocks, while the sedimentaries underlie the lower, gentle river valley floors. This relation holds true for the entire study area, where the upstanding ridges are underlain by harder metamorphics and the valleys mainly by quaternary gravels. Terrain Hypsometry Historically, hypsometry has been used as an indicator of the geomorphic form of the catchments and landforms. Computationally, it refers to finding the distribution of elevations as a function of area occupied by each contour interval within a terrain unit. The idea of hypsometry was first introduced by Langbein and Basil (1947) and was later extended by Strahler (1952). Using dimensionless parameters such

as proportionate area and proportionate altitude, hypsometric curves can be plotted, described and compared irrespective of the true absolute scale. Curves show distinctive differences both in sinuosity of form and in proportionate area below the curve, termed as the hypsometric integral (HI). The geometry of hypsometric curve and the magnitude of the hypsometric integral together describe the stage of evolution of a landscape (Strahler, 1957). Hence, the shape of the curve, the hypsometric integral and the regression parameters of the best-fit statistical curves can be conveniently used as descriptive parameters for the purpose of comparison and classification and to determine the stage of a landscape in the evolutionary process. Based on the H.I., a landscape can be genetically classified in terms of the stage of evolution as youthful (HI > 0.5), mature or in equilibrium (HI = 0.5) or old (HI < 0.5), (Strahler, 1957). The hypsometric curve of the study area (Fig. 9) is steeply concaved upwards – indicative of a landscape in the “old stage” of the normal cycle of erosion. This is further confirmed by the magnitude


21

Fig. 8 Tilted 3-D view of the Duling N-Basin looking Northwest-ward from its mouth and litho-cover of the Duling N-Basin draped over the D.E.M. of the basin.

of hypsometric integral (0.18), implying that of the original landmass of the basin area, only 18% is still left to be denuded away and reduce the entire region into a flat, almost featureless plain. The HI is in accordance with the tectonic and denudational history of the area. The area is tectonically relatively stable with little or no uplift in the recent geologic past. The major stream coursing through the area, the Subarnarekha river, being a superimposed river, has been gouging out the landscape for eons, reducing it to the present, largely gently, undulating state with some remnant uplands still holding out against the ravages of exogenous agents and time. Delineation of terrain units Altitudinal frequency analysis, relief-slope-hillshade analysis, surface profile analysis, 3D visualisations

and finally correlations with the geology enable classifying the study area into distinctive terrain units. To further derive guidelines for this, fractal statistics of the SRTM DEM have been derived using the relevant module in Landserf software. The fractal dimension (FD) has been computed by selecting a fixed interval lag and computing the standard variogram for the surface based on this lag, assuming an isotropic process, and then plotting the log of the lag against the log of the variogram. The FD for the above surface varies between 2 to 3 with a mean value of 2.83. The two extremes are the western higher hills and ridges (FD ~ 2) and the eastern broad valley flats (FD ~ 3). As outlined before, the various terrain units are mainly distinguished on the basis of their differing elevation ranges (summarised from the various ranges

22


Table 2 Demarcated terrain attributes from profile analysis Topographic unit

Sub – divisions

Major Divisions

Attributes

level (m)

1. Erosional Plain 1(a) Gentle Plains

2. Plateau Rim

Altitudinal

below 100

Distinct topographic levels;

1(b) Gently Rolling Plains

100 – 120

Ruggedness rapidly vanishes towar

1(c) Undulating Plains

140 – 160

lower plains; Shallow and wide

1(d) Rugged Plains

200 – 220

valleys; Fewer residual hills.

2(a) Moderately Dissected Zone

360 – 380

Accordant summit levels distinctive;

2(b) Highly Dissected Zone

480 – 500

Open valleys; Rugged terrain.

680 – 700

Accordant summit levels distinctive;

3(b) Highly Dissected Zone

860 – 880

Narrow valleys; Conical Peaks;

3(c) Highest Peaks

above 880

Most rugged terrain

3. Plateau Proper 3(a) Moderately Dissected Zone

other and enable better the eventual terrain unit delineation (Fig. 10). The demarcated units show strong correlation with the geology and slope of the area. The upstanding segments of the terrain are of much less spatial dimension. It is the lower elevation classes that dominate with the gentle plains and dissected lower plateau covering over half the map area (Table 3), suggesting a much denuded landscape with remnant higher hill and residual ridges. Conclusion

Fig. 9 Hypsometric curves and integral for Study area DEM.

estimated earlier) and as such, contours are primarily used for demarcating their extents. Contours for the study area were obtained via digitisation of the constituent topographical maps. Where more sensitive contours were required (lesser contour interval values), these were generated from the SRTM DEM. The two datasets fused together complement each

The SRTM mission achieved what conventional cartography had failed to achieve in three centuries of its existence—to generate a uniform-resolution, uniform-accuracy elevation model of most of the earth’s surface. SRTM DEM usage can facilitate topographic analysis and is a good backup in case topographic maps are not available or that data needs to be enhanced. However, using the SRTM data just for the sake of using it, will lead to dilution of its importance and will lead to ignoring of the much richer geo-spatial content of topographic maps. The results obtained from SRTM grid processing should always be corroborated with available maps or non-spatial data. Despite being a model of continuous surface form, a DEM is a set of discrete elevation true surface will depend on the surface roughness and DEM

23


Fig. 10 Terrain classification of the study area DEM.

Table 3 Demarcated Terrain Class Statistics Land Unit Class

Elevation Class (m)

No. of Cells

Mean Elevation (m)

Elevation Range (m)

Standard Deviation

I

Broad, gentle valley

30 - 100

233076

76.3097

70

15.609

II

Dissected lower plateau

100 - 225

449202

156.799

125

35.0231

III

Dissected higher plateau

225 - 375

284860

284.384

150

40.5975

IV

Steep pediments

375 - 500

57445

426.677

125

34.2011

V

Higher hills

500 - 750

26185

593.507

250

68.2072

VI

Highest peaks

750 - 930

2464

801.12

130

40.3836

24

resolution. Yet, fractal logic suggests that there will always be detail at a finer scale than that measured at the DEM resolution. Thus, all DEMs implicitly model at a certain scale implied by the grid cell resolution (Garbrecht and Martz, 1993; Ackerman, 1993; Hodgson, 1995). Although the scale dependency of measurement has been recognised and even modelled by many authors it still remains hidden in many aspects of DEM analysis. Axiomatically, a higher DEM resolution results in more accurate modelling.

References Ackermann F (1993) Automatic generation of digital elevation models, presented at OEEPE Commission B, DTM Accuracy Meeting, S outhampton. 16 pp Bamler R (1999) The SRTM Mission – A World-Wide 30 m Resolution DEM from SAR Interferometry in 11 Days. Photogrammetric Week 47: 145-154 Burrough PA (1986) Principles of Geographical Information Systems for Land Resources Assessment. OUP, Oxford, Ch.8, Methods of interpolation pp. 147-166 Evans IS (1972) General geomorphometry, derivatives of altitude, and descriptive statistics. in Chorley, R. J. ed. Spatial Analysis in Geomorphology, Methuen, London. pp 17-90 Farr TG and M Kobrick (2000) Shuttle Radar Topography Mission produces a wealth of data, Amer. Geophys. Union Eos, 81 : 583-585 Garbrecht J and Martz L (1993) Grid size dependency of parameters from digital elevation models, 12 pp. (source unknown) Hodgson ME (1995) What cell size does the computed slope/aspect angle represent? Photogrammetric


Engineering and Remote Sensing 61(5) : 513517 Langbein WB and Basil W (1947) Topographic characteristics of drainage basins. USGS WaterSupply Paper, 968-C : 125-158 Rodriguez E, Morris CS, Belz JE, Chapin EC, Martin JM, Daffer W and Hensley S (2005) An assessment of the SRTM topographic products, Technical Report JPL D-31639, Jet Propulsion Laboratory, Pasadena, California, pp. 143 Shuttle Radar Topography Mission DTED ® Level 1 (3-arc second) documentation.http:// edcsns17.cr.usgs.gov/srtm/index.html SRTM Documentation. ftp://edcsgs9.cr.usgs.gov/ pub/data/srtm/Documentation/SRTM_Topo.txt Strahler AN (1952) Hypsometric analysis of erosional topography Geol. Soc. Amer. Bull. 63 : 117- 1142 Strahler AN (1957) Quantitative analysis of watershed geomorphology, Trans Amer Geophys Union. 38 : 913 – 920 Thornbury WD (1954) Principles of Geomorphology, Chapman and Hall, London Weibel R and Heller M (1991) Digital Terrain Modelling In: Maguire, D J., Goodchild, M. F., and Rhind, D. W. (eds.) Geographical Information Systems: Principles and Applications, pp.269-297, Longman, London Werner M (2001) Shuttle Radar Topography Mission (SRTM), Mission overview, J Telecom (Frequenz), 55, 75-79 Wood J (1996) The geomorphological characterisation of digital elevation models Unpublished PhD thesis, Department of Geography, University of Leicester (http://www.soi.city.ac.uk/~jwo/phd) Wood J (2002) Landserf: visualisation and analysis of terrain models (http://www.landserf.org/)