A comparison of SRTM and high-resolution digital elevation models and their use in catchment geomorphology and hydrology: Australian examples

Transcription

1 Earth Surface Processes and Landforms 1394 Earth Surf. Process. Landforms 31, (2006) G. R. Hancock et al. Published online 18 May 2006 in Wiley InterScience ( A comparison of SRTM and high-resolution digital elevation models and their use in catchment geomorphology and hydrology: Australian examples G. R. Hancock, 1 * C. Martinez, 1 K. G. Evans 1,2 and D. R. Moliere 2 1 School of Environmental and Life Sciences, The University of Newcastle, Callaghan, New South Wales 2308, Australia 2 Hydrological and Ecological Processes Program, Environmental Research Institute of the Supervising Scientist, Darwin, Northern Territory, Australia *Correspondence to: G. R. Hancock, School of Environmental and Life Sciences, The University of Newcastle, Callaghan, NSW, 2308, Australia. gggh@alinga.newcastle.edu.au Received 8 March 2005; Revised 20 October 2005; Accepted 16 November 2005 Abstract The recently released Shuttle Radar Topography Mission (SRTM) 3-arc second digital elevation data set provides a complete global coverage of the Earth s land surface. In this paper we examine the SRTM data for three catchments in Australia over a range of climates, geology and resultant geomorphology. To test this new data set the SRTM data are compared with high resolution digital elevation models. We use basic hydrological and geomorphological statistics and descriptors such as the area slope relationship, cumulative area distribution and hypsometric curve, along with Strahler and networking statistics. The above measures describe the surface morphology of a catchment, therefore integrating catchment geology, climate and vegetation. The SRTM data were also assessed as input into the SIBERIA landscape evolution and soil erosion model as were runoff properties, using a wetness index. The results demonstrate that the 90 m SRTM data provide a poor catchment representation. Hillslopes appear as a linked set of facets and display little of the complex curvature that is observed in high resolution data. While catchment area slope and area elevation (hypsometry) properties are largely correct, catchment area, relief and shape (as measured by the width function) are poorly captured by the SRTM data. Catchment networking statistics are also variable. The large grid size of the SRTM data also results in incorrect drainage network patterns and different runoff properties. Consequently, care must be used for quantitative assessment of catchment hydrology and geomorphology, as in all cases SRTM-derived catchment area is incorrect and smaller digital elevation grid sizes are required for accurate catchment-wide assessment. While only a limited number of catchments have been examined, we believe our findings are applicable to other areas. Crown Copyright Reproduced with the permission of the Controller of HMSO. Published by John Wiley & Sons, Ltd. Keywords: digital elevation model; Shuttle Radar Topography Mission; geomorphology; hydrology; geomorphic modelling; hydrological modelling; SIBERIA; soil erosion Introduction Digital elevation models provide a wealth of information regarding catchment geomorphology and hydrology. In recent years many new methods for creating digital elevation models of the Earth s surface have become available, with many new data sets available for public use. Also in the past decade there has been an abundance of studies examining the effect of different digital elevation model grid scales on the ability of a digital elevation model to accurately and reliably represent catchment form and function (e.g. Zhang and Montgomery, 1994; Gyasi-Agyei et al., 1995; Walker and Willgoose, 1999). This study examines the recently released Shuttle Radar Topography Mission (SRTM) 3-arc second digital elevation for its ability to correctly capture catchment geomorphology and terrain-related hydrological variables. To test this new data set the SRTM data are compared with high resolution digital elevation models over a range of catchment

2 Digital elevation models in catchment geomorphology and hydrology 1395 sizes with very different geomorphology. The SRTM data set is unique in that it provides global coverage at a resolution of approximately 90 m by 90 m horizontal grid size. If this data set proves reliable it will provide a new and powerful tool with which to examine the Earth s surface, in our case the interest being hydrological and geomorphological processes. In this study we use basic hydrological and geomorphological statistics and descriptors to examine the SRTM data. The primary focus of this work is not the accuracy of the elevation data themselves, but the attributes derived from them. In previous studies the area slope relationship, cumulative area distribution and hypsometric curve have been used to understand and compare catchments (Hancock and Willgoose, 2001, 2002; Hancock, 2003; Willgoose et al., 2003). These geomorphic descriptors, along with Strahler (1964) and stream networking statistics, are believed to be integrative measures which graphically describe the surface morphology of a catchment, therefore integrating catchment geology, climate and vegetation over geological time. Digital elevation models are also widely used in hydrology and geomorphology. While the above geomorphological descriptors provide an assessment, it is also important that the data be evaluated as landscape inputs into models that use this type of landscape information. While the width function provides a measure of hydrological response (Rodriguez- Iturbe and Valdez, 1979; Naden, 1992), the authors are particularly interested in the use of the SRTM data set as input into hydrological and sediment transport models. For our purposes we will assess the SRTM data as input into the SIBERIA landscape evolution and soil erosion model (Willgoose et al., 1991a,b,c,d) and will provide information regarding the usefulness of the SRTM data for catchment-wide soil erosion assessment. The effect of different digital elevation model grid scale on runoff properties of the catchments will be examined using the Hydrogeomorphic Steady State (HGSS) model of Willgoose and Perera (2001). Wetness indices have been extensively used to assess the runoff properties of catchments (Moore et al., 1991; Wolock and McCabe, 2000). By comparing geomorphological and hydrological descriptors together with an assessment of the SRTM data as model input, the strengths and weaknesses of this new data set are evaluated using a wholecatchment approach. Study Sites The SRTM data are compared to three existing high resolution digital elevation models in different regions of Australia. These catchments cover a range of climates, geology and geomorphology, and catchment areas (Table I, Figures 1 to 3). The locations and characteristics of these catchments are described below. Tin Camp Creek is a natural site undisturbed by Europeans in Arnhem Land, Northern Territory, Australia. The catchment has a geology very similar to the Energy Resources Australia Ranger uranium mine (ERARM); it is thought to be an analogue for the long-term rehabilitated post-mining landscape and has undergone extensive examination in recent years as a result of this (Riley and Williams, 1991; Moliere et al., 2002; Hancock et al., 2002; Hancock, 2003, 2005a,b; Willgoose et al., 2003). The Swift Creek catchment is located approximately 230 km east of Darwin and 20 km northeast of Jabiru in the Northern Territory, Australia. Similar to Tin Camp Creek the site is undisturbed by Europeans. This catchment Table I. Catchment statistics for Tin Camp Creek, Swift Creek and Jemmys Creek Tin Camp Creek Swift Creek Jemmys Creek 90 m 90 m 90 m 10 m 90 m SRTM 20 m 90 m SRTM 100 m SRTM Area (ha) Relief (m) α Hypso. integral Net. convergence Strahler order Bifurcation ratio Slope ratio Length ratio Area ratio

3 1396 G. R. Hancock et al. Figure 1. Tin Camp Creek catchment with a high resolution 10 m by 10 m DEM (top), regridded 90 m by 90 m DEM (middle), and 90 m by 90 m SRTM DEM (bottom). is located in the wet/dry tropics region and has a climate very similar to Tin Camp Creek (Saynor et al., 2004). The catchment is located on the Arnhem Land sandstone plateau and flows to the Magela Creek floodplain. In this study we examine a major tributary of Swift Creek East Tributary. The upper reaches of the study catchment flow through a sandstone-bedrock-confined channel on the plateau. The catchment then flows onto a wooded lowland

4 Digital elevation models in catchment geomorphology and hydrology 1397 Figure 2. Swift Creek catchment with high resolution 20 m by 20 m DEM (top), regridded 90 m by 90 m DEM (middle), and 90 m by 90 m SRTM DEM (bottom).

5 1398 G. R. Hancock et al. Figure 3. Jemmys Creek catchment and drainage network using a 100 m by 100 m DEM (top) and the 90 m by 90 m SRTM DEM (bottom). The box indicates the area of catchment missing from the SRTM data. Horizontal units are metres multiplied by 100 while the vertical dimensions are metres. area which has significantly smaller areas of bedrock. Vegetation species are very similar to those found at Tin Camp Creek. Jemmys Creek is located in the Krui River catchment in the Upper Hunter Valley, New South Wales, Australia. The geology of the area is Tertiary basalt (Atkinson, 1966; Galloway, 1963a,b), a product of Cainozoic volcanism which took place throughout much of eastern Australia (Branagan and Packham, 2000). The site lies within the East Australian Tertiary Volcanic Province, one of four major geologic regions of eastern Australia occupied by the Hunter Valley (Galloway, 1963a).

6 Digital elevation models in catchment geomorphology and hydrology 1399 Digital Elevation Model Data In this study the SRTM data are compared to high resolution digital elevation models (DEMs) for Tin Camp Creek, Swift Creek and Jemmys Creek. The SRTM and other high resolution data used in this study are described below. Shuttle Radar Topography Mission (SRTM) 3-arc second (90 m) DEM The Shuttle Radar Topography Mission (SRTM) 3-arc second DEM is the result of a collaborative effort by the National Aeronautics and Space Administration (NASA), the National Imagery and Mapping Agency (NIMA), the German space agency, and Italian space agency (Rabus et al., 2003; Foni and Seal, 2004; van Zyl, 2001). The mission was launched on 11 February 2000 aboard the Space Shuttle Endeavour. Using radar interferometry, a 3-arc second (SRTM-3) and a 1-arc second (SRTM-1) DEM were produced for almost the entire globe. The Australian SRTM-3 data were publicly released in July 2004, although the SRTM-1 data are yet to be released. Data were collected using two interferometers, C-band (American) and X-band (German) systems, at 1-arc second (30 m) (Foni and Seal, 2004; Rabus et al., 2003; van Zyl, 2001). The absolute vertical and horizontal accuracy of the data collected was reported to be ±16 m and ± 20 m (Kaab, 2005; Kellndorfer et al., 2004; Miliaresis and Paraschou, 2005; Rabus et al., 2003). The 3-arc second (90 m) DEM was created by 3 3 averaging of the 1-arc second data (i.e. 9 data points combined to form a single 3-arc second data point). Elevation data error has features of random noise. Thus the process of averaging is considered to reduce error by approximately a factor of three and reduces random error but not systematic error (USGS, 2003). Each data tile covers an area spanning 1 in latitude and longitude, containing 1201 rows and 1201 columns. Elevation values are given in metres and WGS84 is used as horizontal and vertical datum (Rabus et al., 2003). When corrected to metres, the spacing is m by m, which in this paper is referred to as a 90 m DEM. In this study the Tarboton et al. (1989) method was used to remove all pits in the data. SRTM data tiles used in analysis were reprojected to Universal Transverse Mercator (UTM) for the appropriate and corresponding AMG zone in ERDAS Imagine (v 8.6) software package so as to account for the Earth s curvature. For this study three tiles, Tin Camp Creek (s13e133model), Swift Creek (s12e133model) and Jemmys Creek (s32e150 and s33e150) were used (Figures 1 to 3). Tin Camp Creek and Swift Creek DEMs Tin Camp Creek and Swift Creek landscape data (X,Y,Z coordinates) were determined by digital photogrammetry by AIRESEARCH Pty Ltd, Darwin, and were supplied as irregularly spaced data points within an irregularly shaped boundary. To place the data onto a regular grid, simple kriging (using Surfer v. 7) was used to interpolate the landscape elevation data onto a 10 m by 10 m grid and 20 m by 20 m grid for the Tin Camp Creek and Swift Creek catchments respectively. This 10 m by 10 m and 20 m by 20 m spacing was equivalent to the average spacing of the original AIRESEARCH data over the catchments examined. Further digital elevation models were produced at 90 m by 90 m matching that of the SRTM data using simple kriging (Figures 1 and 2). The Tarboton et al. (1989) method was used to remove all pits. Land and Property Information New South Wales DEM The 25 m digital elevation model was obtained from the Land and Property Information New South Wales (LPI) and was used for comparison with the Jemmys Creek SRTM data. The digital elevation model is in raster format, with a grid spacing of 25 m, and is based on the Australian Map Grid (AMG) coordinate system. Elevation values are provided in metres and referenced to the AUS66 geoid. Further digital elevation models were produced at 100 m by 100 m closely matching that of the SRTM data using a 4 4 averaging (i.e. regridding) of the original 25 m grid LPI data via simple kriging using the commercially available Surfer (v.7) software package (Figure 3). As with the SRTM data, the Tarboton et al. (1989) method was used to remove all pits in the data. Results of Geomorphological Assessment Qualitative or visual assessment of catchment morphology has been shown to provide an important first step in assessing and comparing catchments (Hancock et al., 2002; Hancock, 2003). If a catchment is visually different from a neighbour that is subject to the same climate then it is likely to be geomorphologically quantitatively different (Hancock, 2003). In this paper the catchments are assessed using both qualitative (visual) and quantitative measures.

7 1400 G. R. Hancock et al. Quantitative measures used are the area slope relationship, hypsometric curve, cumulative area distribution and the width function. Other quantitative measures such as network statistics, catchment energy and runoff generation were also used for the geomorphological assessment. The area slope relationship is the relationship between the area draining through a point versus the slope at that point. It quantifies the local topographic gradient as a function of drainage area. Two distinct regions of the relationship are typically observed. Small catchment areas are dominated by rainsplash, interrill erosion, soil creep or other erosive processes that tend to round or smooth the landscape. As the catchment area becomes larger, a break in gradient of the curve occurs. This is where slope decreases as catchment area increases. This region of the catchment is dominated by fluvial erosive processes, that is, those processes that tend to incise the landscape. A relationship of the form A α S = constant (1) where A is the contributing area to the point of interest, and S is the slope of the point of interest, is common. This equation applies to the fluvial (log log linear) part of the relationship, where α is the slope of the fluvial region of the area slope relationship. This break in slope in the data can be determined by qualitative assessment and/or by using the break in slope for the cumulative area distribution (discussed below) which clearly delineates this change. The area slope relationship is considered to be a fundamental geomorphic relationship with the value of α ranging between 0 4 and 0 7 for natural catchments (Hack, 1957; Flint, 1974; Gupta and Waymire, 1989; Tarboton et al., 1992; Montgomery and Foufoula-Georgiou, 1993; Willgoose, 1994; Montgomery and Dietrich, 1988, 1989, 1994). The hypsometric curve (Langbein, 1947) is a non-dimensional area elevation curve which allows a ready comparison of catchments with different area and steepness. The hypsometric curve has been used as an indicator of the geomorphic maturity of catchments and landforms (Strahler, 1952, 1964). Strahler (1952, 1964) divided landforms into young, mature and monadnock characteristic shapes, reflecting increasing catchment age. Willgoose and Hancock (1998) demonstrated that these characteristic shapes were also consistent with different catchment erosion processes, catchment geometry and network form. The cumulative area distribution is a function defining the proportion of the catchment that has a drainage area greater than or equal to a specified drainage area, and describes the spatial distribution of areas and drainage network aggregation properties within a catchment. The cumulative area distribution has been used as a means of characterizing the flow aggregation structure of channel networks (Rodriguez et al., 1992; LaBarbera and Roth, 1994). The cumulative area distribution is similar to the area slope relationship in that it provides the ability to examine the relationship between diffusive and fluvial processes. Similar to the hypsometric curve, the cumulative area distribution is indirectly related to the area slope relationship as the distribution of areas in a catchment is related to its area elevation properties. The cumulative area distribution typically has a form that consists of three regions. Region one represents those small areas of the catchment where rainsplash or interrill erosion is the dominant erosive mechanism, and it is also the region of hillslope flow aggregation. This region is largely influenced by diffusive erosion processes (Hancock and Willgoose, 2001; Hancock et al., 2002). Region two represents catchment area dominated by channelized flow. This region is generally observed to be approximately log log linear and is influenced by the value of α in the area slope relationship (Willgoose, 1994). Region three consists of that part of the catchment dominated by large channels near the catchment outlet. Large area contributions are made in this part of the catchment where the ordinate of the distribution function rapidly decreases as a result of increasing drainage area. Descriptors of channel networking properties used here are the width function, network convergence and Strahler statistics. The width function (Surkan, 1968) is a plot of the number of channels at a given distance from the basin outlet, measured along the network (Naden, 1992). A slightly more general interpretation is adopted here, which is easier to apply for digital terrain maps. The width function used here is the number of drainage paths (whether they be channel or hillslope) at a given distance from the outlet as it is difficult to determine what is channel and what is hillslope on a digital terrain map. Catchment drainage network convergence for a gridded digital terrain map is the average number of channels draining into a point in a catchment. Convergence statistics provide, in addition to the width function, a further method of analysing catchment drainage and network properties (Perera and Willgoose, 1998; Ibbitt et al., 1999). Qualitative assessment The 10 m grid digital elevation model for Tin Camp Creek displays well-rounded hillslope of regular curvature and hillslope length over the entire domain, and is well dissected by a regularly spaced drainage network (Figure 1).

8 Digital elevation models in catchment geomorphology and hydrology 1401 Figure 4. Area slope relationship for Tin Camp Creek (top), Swift Creek (middle) and Jemmys Creek (bottom). For clarity the 10 m, 20 m and 90 m SRTM digital elevation model data for the slope are divided by 10. Swift Creek, using the 20 m data, appears as a much more dissected landscape reflective of its incised sandstone bedrock (Figure 2). Visual examination of the catchments with coarser grid spacing (as compared to the finer spacing) demonstrates a loss of surface morphological detail at the larger grid scales. Using the SRTM and 90 m data the catchment appears as a set of linked linear facets with hillslope curvature being particularly poorly represented. Much of the hillslope and channel detail has been lost and subsequently a considerable amount of catchment area has been lost (Table I). Catchment area for the SRTM data is considerably lower than the other data sets for Tin Camp Creek, while the reverse applies for Swift Creek. In all cases, relief is higher for the 10 m and 20 m data than for the low resolution data. Figure 3 demonstrates that there is little visual difference when comparing the two data sets for Jemmys Creek and that loss of catchment area is not simply related to the location of the catchment outlet. Area slope relationship The area slope relationship using high resolution data displays a tightly grouped data set in both the diffusive (concave) and log log linear (fluvial) erosion-dominated sections of the curve for Tin Camp and Swift Creek (Hancock et al., 2002) (Figure 4). The data sets have a diffusive region of approximately 10 pixels, the slope of the fluvially

9 1402 G. R. Hancock et al. dominated region being 0 34, 0 43 and 0 47 for Tin Camp, Swift Creek, and Jemmys Creek, respectively, and are largely log log linear (Hancock et al., 2002) (Table I). The slope (or α value) of the fluvial region of the data is typical of fluvially dominated landscapes. An examination of the area slope data for the different grid spacing demonstrates that as grid size increases, detail in the area slope relationship is lost (Figure 4). At large grid spacings the curvature in the diffusive region is lost and the area slope relationship is log log linear for its entire domain. The slope (α value) of the log log linear section of the curve also varies with a large difference in the Tin Camp Creek data set between the 10 m and regridded 90 m digital elevation models (Table I). Alpha values for the SRTM data sets are slightly higher for Tin Camp Creek and slightly lower for Swift Creek when compared to the high resolution data. Examination of the area slope results for Jemmys Creek suggest there is little difference between the 100 m and 90 m SRTM digital elevation models, with both sets displaying distinctive diffusive and fluvial regions and having similar α values (Figure 4). Hypsometric curve Comparison of the hypsometric curve for the catchments over the different grid scales demonstrates that both the 90 m and SRTM data sets are similar (Table I) to the high resolution data and all have the same shape (Figure 5). This Figure 5. Hypsometric curve for Tin Camp Creek (top), Swift Creek (middle) and Jemmys Creek (bottom).

10 Digital elevation models in catchment geomorphology and hydrology 1403 suggests that the distribution of area and elevation in a catchment is scale invariant and that the hypsometric curve is largely insensitive to grid scale (Hancock, 2005b). The hypsometric curve provides little information on the loss of hillslope and channel definition as grid size increases. Differences occur at the toe of the curve for Swift Creek where the catchment descends from the sandstone plateau onto the sandy floodplain. In this region of the catchment, slopes are low and it is difficult for a coarse-grid digital elevation model to correctly delineate drainage lines and associated drainage areas. Cumulative area distribution The high resolution data for the cumulative area distribution displays three regions typical of that observed for other fluvially dominated catchments (Hancock et al., 2002) (Figure 6). This is highlighted when examining the slope of the cumulative area distribution (Figure 6, bottom). For the high resolution data at areas approximately less than 10, the data have convex curvature representing diffusive-dominated areas of the hillslope, and at areas greater than 10 data display log log linear behaviour representing catchment areas dominated by fluvial processes followed by a break in slope as large catchment areas congregate. For the 90 m and SRTM data sets, considerable detail is lost in the diffusive region of the curve. Similar results were found for both Swift Creek and Jemmys Creek and are not displayed. Channel networking properties The width function displays considerable spatial variability as a result of both catchment shape and catchment size (for comparison the data were normalized by dividing distance and width, by maximum distance and width, respectively) (Figure 7). For Tin Camp Creek the 90 m data set follows the high resolution data for the rising limb but provides a poor match for the falling limb, whereas the 90 m SRTM data set is the opposite. For Swift Creek the rising Figure 6. Cumulative area distribution (top) and slope of the cumulative area distribution (bottom) for Tin Camp Creek.

11 1404 G. R. Hancock et al. Figure 7. Normalized width function for Tin Camp Creek (top), Swift Creek (middle) and Jemmys Creek (bottom). limb of the SRTM data is higher than that of the 20 m and 90 m data sets. Nevertheless the falling limb is well matched. For the non-normalized width function for Jemmys Creek (not displayed), while the overall shape of the width functions are comparable for the two DEMs, the 100 m DEM width function is slightly shorter in distance and narrower than that recorded for the 90 m DEM, reflecting the slightly different grid sizes. Normalizing the data (Figure 7) produces very similar width functions. Convergence statistics (from the drainage network determined using the D8 algorithm; Table I) demonstrate that the coarse grid and SRTM data have a higher network convergence than the high resolution 10 m digital elevation model catchments and that the 10 m digital elevation model has a less branched network than the 90 m and 90 m SRTM digital elevation models. This demonstrates that as catchment grid spacing increases so does the network convergence value, suggesting that grid size has an effect on drainage network characterization. As the catchment grid spacing increases, the maximum Strahler stream order of the catchments decreases while the slope, length and area ratios increase (Table I). The bifurcation ratio for the SRTM data approximates the high resolution data whereas the coarse grid data are consistently higher than both the other data sets. This demonstrates that choice of digital elevation model grid spacing has a direct impact on Strahler network properties and is reflecting the more branched network as grid spacing increases.

12 Digital elevation models in catchment geomorphology and hydrology 1405 Evaluation of the 10 m, 90 m and 90 m SRTM Digital Elevation Models in a Landscape Evolution Model and for Catchment Hydrology It is important that the SRTM data set be evaluated as input into models that use digital elevation model data. Digital elevation models are extensively used not only for hydrological assessment but also for sediment transport and soil erosion assessment. In this paper we compare erosion from the 10 m, 90 m and 90 m SRTM digital elevation models when input into the SIBERIA erosion model (Willgoose et al., 1991a,b,c,d), providing a further evaluation of the different data sets. In this study we evaluate the Tin Camp Creek digital elevation model, as a reliable erosion model parameter data set exists for the site. These data have been extensively used in past studies (Hancock et al., 2002; Moliere et al., 2002; Hancock, 2003, 2005a,b; Willgoose et al., 2003). The SIBERIA erosion model and its calibration are discussed below. Runoff properties and the SRTM data determined using the Hydrogeomorphic Steady State (HGSS) model of Willgoose and Perera (2001) are also discussed below. The SIBERIA landscape evolution model SIBERIA is a physically based mathematical model that simulates the geomorphic evolution of landforms subjected to fluvial and diffusive erosion and mass transport processes. SIBERIA links widely accepted hydrology and erosion models under the action of runoff and erosion over long time scales. SIBERIA is an important tool in the understanding of the interactions between geomorphology, erosion and hydrologic processes because of its ability to explore the sensitivity of a system to changes in physical conditions, without many of the difficulties of identification and generalization associated with the heterogeneity encountered in field studies. The sediment transport equation of SIBERIA is q s = q sf + q sd (2) where q s (m 3 s 1 m 1 width) is the sediment transport rate per unit width, q sf is the fluvial sediment transport term and q sd is the diffusive transport term (both m 3 s 1 m 1 width). The fluvial sediment transport term (q sf ), based on the Einstein Brown equation, models incision of the land surface and can be expressed as m1 n1 q = β 1 q S (3) sf where q is the discharge per unit width (m 3 s 1 m 1 width), S (m m 1 ) is the slope in the steepest downslope direction and β 1, m 1 and n 1 are calibrated parameters. The diffusive term, q sd, is q sd = DS (4) where D (m 3 s 1 m 1 width) is diffusivity and S is slope. The diffusive term models smoothing of the land surface and combines the effects of creep, rainsplash and landsliding. SIBERIA does not directly model runoff (Q, m 3, for the area draining through a point) but uses a sub-grid effective parameterization based on empirical observations and justified by theoretical analysis which conceptually relates discharge to area (A) draining through a point as Q = β A m 3 3 (5) where β 3 is the runoff rate constant and m 3 is the exponent of area, both of which require calibration for the particular field site. For long-term elevation changes it is convenient to model the average effect of the above processes with time. Accordingly, individual events are not normally modelled but rather the average effect of many aggregated events over time. As a result, SIBERIA describes how the catchment is expected to look, on average, at any given time. The sophistication of SIBERIA lies in its use of digital terrain maps for the determination of drainage areas and geomorphology, and also its ability to efficiently adjust the landform with time in response to the erosion that occurs on it. A more detailed description of SIBERIA can be found in Willgoose et al. (1991a,b,c,d).

13 1406 G. R. Hancock et al. Calibration of SIBERIA input parameters. Before SIBERIA can be used to simulate soil erosion, the sediment transport equation (Equation 3) and area discharge relationship (Equation 5) require calibration. The fluvial sediment transport equation (Equation 3) in SIBERIA is parameterized using input from field sediment transport and hydrology data. This parameterization process is described in detail by Evans et al. (1999, 2000) and Hancock et al. (2000). For this study the SIBERIA model was calibrated from field data collected at Tin Camp Creek from a series of rainfall events. The calibration of SIBERIA for Tin Camp Creek is described in detail elsewhere (Moliere et al., 2002). A summary is provided below. Two catchments of size 2032 m 2 and 2947 m 2 with average slopes of 19 per cent and 22 per cent, respectively, were instrumented during the wet season of Both sites are incised and channelized and are representative of the area. The study sites were monitored during rainfall events from December At this time the catchments had a good covering of speargrass, which quickly regenerates each wet season. To calibrate the erosion and hydrology models, complete data sets of sediment loss, rainfall and runoff for nine discrete rainfall events in both catchments were collected, allowing calibration for the two individual catchments. The rainfall runoff monitoring data were used to parameterize the DISTFW (Field and Williams, 1983) rainfall runoff model. The parameterized model was used to derive long-term average parameters for SIBERIA. The parameters of SIBERIA represent temporal average properties of the runoff and erosion processes occurring on the landscapes. The parameter values derived above for the DISTFW model represent instantaneous values and must be integrated over time to yield temporal average values. Equation 5 was fitted using the peak discharges and areas for both catchments. Storms of various duration for a 1- in-2-year average return interval (ARI) were used, using rainfall data from Jabiru which has the closest long-term data. The storm durations were 10, 15, 20, 25, 30, 45 and 60 minutes. The ARI of the storm that most closely relates the instantaneous erosion with long-term erosion is 2 33 years (Willgoose et al., 1989). Since the ARI is used solely to determine the value of m 3 in Equation 6, use of the 1-in-2-year rather than 2 33 year is considered satisfactory and consistent with the index flood approach to flood frequency analysis. To determine the sediment transport parameters, Equation 3 can be rearranged and fitted to the total sediment loss during the rainfall events for each catchment (Evans et al., 1998). Willgoose and Riley (1998) used rainfall simulation data on cap and batter site plots at ERARM to derive a value of the slope parameter of n 1 = This value has been shown to be a reliable exponent on slope in this study area (Hancock et al., 2000, 2002). Using this value, m 1 = 1 70 and β 1 = 1067, 234 and 231 (for the 10 m, 90 m and 90 m SRTM digital elevation models, respectively) parameter values were determined (Moliere et al., 2002). SIBERIA simulations. The SIBERIA simulations were run using the Tin Camp Creek 10 m, 90 m and 90 m SRTM digital elevation models and the calibrated erosion parameters above. The simulations were run for a period of 10 years and the results assessed for average, minimum and maximum depths of erosion (Table II). This provides an areaindependent measure of soil loss in the catchments. Average erosion is largely similar, with the 10 m and 90 m catchments producing values of m and m respectively. The 90 m SRTM digital elevation model produced an average erosion rate of m, approximately half that of the other data sets (Table II). This equates to approximately 7 7, 6 9 and 3 6 t ha 1 a 1 for the catchments. Hancock (unpublished data) has found that soil erosion in the area ranges from 2 9 to 14 1 t ha 1 a 1, suggesting that all digital elevation model data sets are providing an adequate assessment of soil loss over the study area. While minimum erosion was zero for the three data sets, there were large differences in maximum erosion with the 10 m, 90 m and 90 m SRTM digital elevation models having a depth of m, m and m respectively (Table II). Position of erosion was different (Figure 8), with the 10 m digital elevation model having the majority of erosion occurring in the major drainage lines in the higher reaches of the catchment, while the 90 m and 90 m SRTM data sets have erosion occurring in the lower reaches of the catchment over a much broader area. Field observation and measurement demonstrate that the 10 m data set provides a more reliable representation of erosion in the catchment. Table II. Average, minimum and maximum depth of erosion for the 10 m, 90 m and 90 m SRTM digital elevation models for Tin Camp Creek 90 m 10 m 90 m SRTM Average (m) Minimum (m) Maximum (m)

14 Digital elevation models in catchment geomorphology and hydrology 1407 Figure 8. Depth (m) and position of erosion at Tin Camp Creek for the SIBERIA simulations using 10 m (top), 90 m (bottom left) and 90 m SRTM (bottom right) digital elevation models. The 10 m digital elevation model is displayed with a drainage support area of 5 pixels and is scaled up for clarity. This figure is available in colour online at Assessment of catchment hydrology for the 10 m, 90 m and 90 m SRTM digital elevation models Wetness indices have been extensively used to assess the runoff properties of catchments (Moore et al., 1991; Wolock and McCabe, 2000). In order to assess the impact of different digital elevation model grid scale on runoff properties of

15 1408 G. R. Hancock et al. the catchments, the Hydrogeomorphic Steady State (HGSS) model of Willgoose and Perera (2001) was used. This model incorporates catchment organization and geomorphological relations of the area slope relationship and the cumulative area distribution (Willgoose and Perera, 2001). In this study the wetness index, λ i, is used to describe the relative area of the catchment saturated when water saturates the total depth of the soil profile. The saturation of the soil profile and resultant saturation excess runoff is a typical runoff process in the monsoonal tropical climate of the Northern Territory. The wetness index is calculated by λ = i 1+ α A i C (6) where A i is catchment area draining through a point, α and C are the exponent and constant, respectively, in the area slope relationship (Equation 1) for the entire catchment. In this case α and C are taken from the 10 m, 90 m and 90 m SRTM data sets for Tin Camp Creek (Table I). The wetness index was overlaid on the drainage network for each catchment (Figure 9). The results demonstrate that the 10 m data produce saturated areas along the major drainage lines (areas that would be expected to be saturated each season) but for the 90 m and 90 m SRTM data sets a much greater area of the catchment is saturated. Consequently, use of this coarse grid size data is likely to result in an overestimate of runoff in these catchments. Discussion The ability to understand catchment processes is reliant on the digital elevation model scale and reliability of landscape data input (Kenward et al., 2000; Thompson et al., 2001; McMaster, 2002). Advances in numerical models to monitor and predict hydrology and geomorphology rely heavily on digital elevation models and their integrity, yet the strengths and limitations of these data sets have not been fully investigated or understood (Moore et al., 1991; Moore and Grayson, 1991; Zhang and Montgomery, 1994; Walker and Willgoose, 1999; Fryer et al., 1994; Lane et al., 1994; Gyasi-Agyei et al., 1995; Quinn et al., 1995; Willgoose and Perera, 2001). New spatial data sets (such as the SRTM digital elevation model) require detailed assessment before they can be used reliably. This study has compared high resolution digital elevation data with the much coarser resolution SRTM data over a range of catchments of different sizes and geomorphology. The results demonstrate that for the smaller catchments the 90 m SRTM data and high resolution data degraded to 90 m and 100 m provide a poor catchment representation. Hillslopes appear as a linked set of facets and display little of the complex curvature that is observed in high resolution data. While catchment area slope and area elevation (hypsometry) properties are similar, catchment area, relief and shape (as measured by the width function) are poorly captured by the SRTM data and high resolution data degraded to 90 m. Catchment networking statistics are also variable. This finding, together with the larger area likely to be saturated in the large grid size digital elevation models, raises questions about the viability of the SRTM data for hydrological modelling. Therefore, caution must be used when applying this data set. A major concern of this study is the inability of this coarse resolution data to correctly capture catchment area. Large differences occur between the high resolution and the SRTM data. This is to be expected as the 90 m SRTM data are not sufficiently fine to capture critical stream junctions, which leads to loss of stream lines and resultant area. Therefore, if a junction is missed the contributing area is missed. The results of the soil erosion assessment demonstrate that the SRTM data set, when input to the SIBERIA erosion model using calibrated erosion parameters, provides lower rates of erosion compared to the high resolution and high resolution degraded to 90 m data. The 90 m data provide a higher sediment output than the SRTM data as the initial photogrammetric process better captured the major topographic features, such as catchment divides and drainage lines (examination of the original ungridded points from the Tin Camp Creek digital elevation shows that the digital photogrammetric process used by AIRESEARCH has preferentially captured the stream lines and catchment boundary very well) (Hancock, 2005c). This low sediment output from the SRTM data is likely to result from the combination of lower catchment relief and smaller catchment area as well as the large pixels producing a poor representation of the major drainage lines. Therefore, as input into landscape process models the SRTM data should be employed with caution. While the diffusive region of the area slope relationship is largely lost for the SRTM data, the fluvial region of the area slope relationship is similar to the high resolution data. This suggests that the SRTM data set may be useful in broad scale derivation and assessment of hydrological and erosion model parameters. Willgoose (1994) theoretically showed that the slope of the area slope relationship (α) can be related to erosion model parameters in the Einstein Brown sediment transport equation. Hancock et al. (2002) demonstrated that α values can be used to calibrate erosion

16 Digital elevation models in catchment geomorphology and hydrology 1409 Figure 9. Drainage network and wetness index for Tin Camp Creek for the 10 m (top), 90 m (bottom left) and 90 m SRTM (bottom right) digital elevation models. This figure is available in colour online at model parameters in the SIBERIA (Willgoose et al., 1991a,b,c,d) landscape evolution model. Therefore the SRTM digital elevation model may be used to determine erosion process and as a tool for the calibration of sediment transport models in ungauged basins. Consequently the results of this study demonstrate that for broad-scale qualitative assessment of large catchments, the SRTM data are of benefit, yet for quantitative assessment of catchment hydrology and geomorphology care must be used. Conclusion The ability to understand catchment processes is reliant on the scale and reliability of available landscape data. Digital elevation models provide the ability to qualitatively and quantitatively analyse the Earth s surface. By comparing

17 1410 G. R. Hancock et al. geomorphological and hydrological descriptors, together with an assessment of the SRTM data as model input, the strengths and weaknesses of this new data set are evaluated using a whole-catchment approach. A major issue is the coarse resolution of the SRTM data. This results in catchments that appear as a linked set of facets which display little of the complex curvature that is observed in high resolution data. A further concern is the inability of this coarse resolution data to correctly capture catchment area. Large differences occur between the high resolution and the SRTM data. This is to be expected as the 90 m SRTM data are not sufficiently fine to capture critical stream junctions, which leads to loss of stream lines and resultant area. Consequently, if a junction is missed the contributing area is missed. These two factors combine to produce catchments which have very different geomorphological and hydrological characteristics than what is expected to occur in the field. The results demonstrate that for broad-scale qualitative assessment of large catchments the SRTM data are of benefit. Smaller grid size digital elevation model data are required for a reliable catchment-wide assessment. Care must be used for quantitative assessment of catchment hydrology and geomorphology, and if employed for modelling purposes scaling properties must be well understood and the model correctly parameterized. Acknowledgements The traditional owners of the land in the Northern Territory where the study site is located, Parks Australia North, The Northern Land Council and Supervising Scientist Group staff, especially John Lowry are thanked for their cooperation and assistance. The support of Landholders in the Merriwa region is also gratefully acknowledged. The advice and support of Olivier Rey-Lescure (digital elevation model preparation), Garry Willgoose together with review comments by David O Brien are very much appreciated. Stuart Lane and a second anonymous reviewer are thanked for their constructive reviews. References Atkinson W A Review of the Land Systems of the Hunter Valley, N.S.W. The Hunter Valley Research Foundation: Newcastle, Australia. Branagan DF, Packham GH Field Geology of New South Wales (third edition). New South Wales Department of Mineral Resources: Sydney, Australia. Evans KG, Willgoose GR, Saynor MJ, House T Effect of vegetation and surface amelioration on simulated landform evolution of the post-mining landscape at ERA Ranger Mine, Northern Territory. Supervising Scientist Report 134. Supervising Scientist: Canberra. Evans KG, Saynor MJ, Willgoose GR Changes in hydrology, sediment loss and microtopography of a vegetated mine waste rock dump impacted by fire. Land Degradation and Development 10: Evans KG, Willgoose GR, Saynor MJ, Riley SJ Post-mining landform evolution modelling. I. Derivation of sediment transport model and rainfall-runoff model parameters. Earth Surface Processes and Landforms 25(7): Field WG, Williams BJ A generalised one-dimensional kinematic catchment model. Journal of Hydrology 60: Flint JJ Stream gradient as a function of order, magnitude and discharge. Water Resources Research 10(5): Foni A, Seal D Shuttle Radar Topography Mission: an innovative approach to shuttle orbital control. Acta Astronautica 54: Fryer JG, Chandler JH, Cooper MAR Short communication on the accuracy of heighting from aerial photographs and maps: Implications for process modelers. Earth Surface Processes and Landforms 19: Galloway RW. 1963a. Part IV: Geology of the Hunter Valley. In General Report on the Lands of the Hunter Valley, Story R, Galloway RW, van de Graaff RHM, Tweedie AD. Land Research Series No. 8. Commonwealth Scientific and Industrial Research Organization, Australia: Melbourne, Australia. Galloway RW. 1963b. Part V: Geomorphology of the Hunter Valley. In General Report on the Lands of the Hunter Valley, Story R, Galloway RW, van de Graaff RHM, Tweedie AD. Land Research Series No. 8. Commonwealth Scientific and Industrial Research Organization, Australia: Melbourne, Australia. Gupta VK, Waymire E On the formation of an analytic approach to hydrologic response and similarity at the basin scale. Journal of Hydrology 65: Gyasi-Agyei Y, Willgoose GR, DeTroch FP Effect of vertical resolution and map scale of digital elevation model on geomorphological parameters used in hydrology. Hydrological Processes 9: Hack JT Studies of longitudinal stream profiles in Virginia and Maryland. United States Geological Survey Professional Paper 292(B): Hancock GR The effect of catchment aspect ratio on geomorphological descriptors. In Prediction in Geomorphology, Wilcock P, Iverson R (eds). Geophysical Monograph Series Volume 135. American Geophysical Union: Washington. DOI: /135GM015. Hancock GR. 2005a. Digital elevation model error and its effect on modelling soil erosion and catchment geomorphology. In Sediment Budgets II (Proceedings of symposium S1 held during the Seventh IAHS Scientific Assembly at Foz do Iguaçu, Brazil, April 2005). IAHS Publication 292. Hancock GR. 2005b. The use of digital elevation models in the identification and characterization of catchments over different grid scales. Hydrological Processes 19: Hancock GR. 2005c. The impact of different gridding methods on catchment geomorphology and soil erosion over long time scales using a landscape evolution model. Earth Surface Processes and Landforms. DOI:10/1002/esp.1306.

18 Digital elevation models in catchment geomorphology and hydrology 1411 Hancock GR, Willgoose GR The use of a landscape simulator in the validation of the SIBERIA landscape evolution model, steady state landforms. Water Resources Research 37(7): Hancock GR, Willgoose GR The use of a landscape simulator in the validation of the SIBERIA landscape evolution model: transient landforms. Earth Surface Processes and Landforms 27: Hancock GR, Willgoose GR, Evans KG, Moliere DR, Saynor MJ Medium term erosion simulation of a abandoned mine site using the SIBERIA landscape evolution model. Australian Journal of Soil Research 38: Hancock GR, Willgoose GR, Evans KG Testing of the SIBERIA landscape evolution model using the Tin Camp Creek, Northern Territory, Australia, field catchment. Earth Surface Processes and Landforms 27(2): Ibbitt RP, Willgoose GR, Duncan MJ Channel network simulation models compared with data from the Ashley River, New Zealand. Water Resources Research 35(12): Kaab A Combination of SRTM3 and repeat ASTER data for deriving alpine glacier flow velocities in the Bhutan Himalaya. Remote Sensing of Environment 94: Kellndorfer J, Walker W, Pierce L, Dobson C, Fites JA, Hunsaker C, Vona J, Clutter M Vegetation height estimation form Shuttle Radar Topography Mission and National Elevation Datasets. Remote Sensing of Environment 93: Kenward T, Lettenmaier DP, Wood EF, Fielding E Effects of digital elevation model accuracy on hydrologic predictions. Remote Sensing of Environment 74: LaBarbera P, Roth G Invariance and scaling properties in the distributions of contributing area and energy in drainage basins. Hydrological Processes 8: Lane SN, Chandler JH, Richards KS Developments in monitoring and modelling small scale river bed topography. Earth Surface Processes and Landforms 19: Langbein WB Topographic characteristics of drainage basins. US Geological Society Water Supply Paper 968-C. Washington, DC. McMaster KJ Effects of digital elevation model resolution on derived stream network positions. Water Resources Research 38(4): Miliaresis GC, Paraschou CVE Vertical accuracy of the SRTM DTED level 1 of Crete. International Journal of Applied Earth Observation and Geoinformation 7: Moliere DR, Evans KG, Willgoose GR, Saynor MJ Temporal trends in erosion and hydrology for a post-mining landform at Ranger Mine. Supervising Scientist Report 165. Supervising Scientist: Darwin, NT. Montgomery DR, Dietrich WE Where do channels begin? Nature 336: Montgomery DR, Dietrich WE Source areas, drainage density and channel initiation. Water Resources Research 25(8): Montgomery DR, Dietrich WE Landscape dissection and drainage area slope thresholds. Process Models and Theoretical Geomorphology, Kirkby MJ (ed.). John Wiley and Sons: Chichester: Montgomery DR, Foufoula-Georgiou E Channel network source presentation using digital elevation models. Water Resources Research 29(12): Moore ID, Grayson RB Terrain-based catchment partitioning and runoff prediction using vector elevation data. Water Resources Research 27(6): Moore ID, Grayson RB, Ladson AR Digital terrain modelling: A review of hydrological, geomorphological and biological applications. Hydrological Processes 5: Naden PS Spatial variability in flood estimation for large catchments: the exploitation of channel network structure. Journal of Hydrological Sciences 37(1): Perera HJ, Willgoose GR A physical explanation of the cumulative area distribution curve. Water Resources Research 34(5): Quinn PF, Beven KJ, Lamb R The ln(a/tanb) index. How do you calculate it and how to use it within the TOPMODEL framework. Hydrological Processes 9: Rabus B, Eineder M, Roth A, Bamler R The shuttle radar topography mission a new class of digital elevation models acquired by spacebourne radar. ISPRS Journal of Photogrammetry and Remote Sensing 57: Riley SJ, Williams DK Thresholds of gullying, Arnhem Land, Northern Territory. Malaysian Journal of Tropical Agriculture 22(2): Rodriguez-Iturbe I, Valdes JB The geomorphologic structure of hydrologic response. Water Resources Research 15: Rodriguez I, Ijjasz Vasquez EJ, Bras RL, Tarboton DG Distributions of discharge, mass and energy in river basins. Water Resources Research 28(4): Saynor MJ, Erskine WD, Evans KD, Eliot I Gully initiation and implications for management of scour holes in the vicinity of Jabiluka Mine, Northern Territory, Australia. Geografiska Annaler 86(A): Strahler AN Hypsometric (area altitude) analysis of erosional topography. Geological Society of America Bulletin 63: Strahler AN Quantitative geomorphology of drainage basins and channel networks. In Handbook of Applied Hydrology, Chow VT (ed.). McGraw-Hill: New York; Surkan AJ Synthetic hydrographs: Effects of network geometry. Water Resources Research 5(1): Tarboton DG, Bras RL, Rodriguez-Iturbe I The analysis of river basins and channel networks using digital terrain data. Technical Report 326. Department of Civil and Environmental Engineering, Massachusetts Institute of Technology: Boston. Tarboton DG, Bras RL, Rodriguez-Iturbe I A physical basis for drainage density. Geomorphology 5(1/2): Thompson JA, Bell JC, Butler CA Digital elevation model resolution: effects on terrain attribute calculation and quantitative soillandscape modelling. Geoderma 100: