Delineating soil landscape facets from digital elevation models using compound topographic index in a geographic information system.
Abstract: Soil landscapes and their component facets (or sub-units) are fundamental information for land capability assessment and land use planning. The aim of the study was to delineate soil landscape facets from readily available digital elevation models (DEM) to assist soil constraint assessment for urban and regional planning in the coastal areas of New South Wales (NSW), Australia. The Compound Topographic Index (CTI) surfaces were computed from 25 m DEM using a D-infinity algorithm. The cumulative frequency distribution of CTI values within each soil landscape was examined to identify the values corresponding to the area specified for each unmapped facet within the soil landscape map unit. Then these threshold values and CTI surfaces were used to generate soil landscape facet maps for the entire coastal areas of NSW. Specific programs were developed for the above processes in a geographic information system so that they are automated, fast, and repeatable. The modelled facets were assessed by field validation and the overall accuracy reached 93%. The methodology developed in this study has been proven to be efficient in delineating soil landscape facets, and allowing for the identification of land constraints at levels of unprecedented detail for the coast of NSW.

Additional keywords: soil landscape, compound topographic index, terrain modelling, geographic information system, GIS.
Article Type: Report
Subject: Geographic information systems (Methods)
Land use (Planning)
Land use (Research)
Soil surveys
Authors: Yang, X.
Chapman, G.A.
Gray, J.M.
Young, M.A.
Pub Date: 12/01/2007
Publication: Name: Australian Journal of Soil Research Publisher: CSIRO Publishing Audience: Academic Format: Magazine/Journal Subject: Agricultural industry; Earth sciences Copyright: COPYRIGHT 2007 CSIRO Publishing ISSN: 0004-9573
Issue: Date: Dec, 2007 Source Volume: 45 Source Issue: 8
Topic: Event Code: 310 Science & research Computer Subject: Geographic information system
Product: SIC Code: 7372 Prepackaged software
Geographic: Geographic Scope: New South Wales Geographic Code: 8AUNS New South Wales
Accession Number: 174059452
Full Text: Introduction

This study formed part of the Comprehensive Coastal Assessment (CCA) project initiated by New South Wales (NSW) Government in 2004. The NSW Coastal Strategy formally recognised that the NSW coast is undergoing rapid population growth, demographic change, and land use intensification. A key to managing changes and ensuring orderly sustainable development along the NSW coast is to ensure that all land is used according to its capability. Land capability may be defined as the inherent physical potential of the land to effectively support a land use without degradation to land and water resources (Dent and Young 1981).

Maps presenting specific land use potential, limitations, or hazards, known as capability or constraint maps, can be used to identify areas of high land use versatility--land which is capable of being used for many different purposes. If land is used according to its capability, the potential for land use conflicts can be greatly reduced. Soil landscape maps and soil attributes are essential information for such land capability assessment.

Soil landscapes are defined as 'areas of land that have recognisable and specifiable topographies and soils, that are capable of presentation on maps and can be described by concise statements' (Northcote 1978). The mapping of landscape properties can be used to distinguish mappable areas of soils because similar causal factors (such as climate, topography, and parent materials) are involved in the formation of both landscapes and soils. This allows both soil and landscape limitations and other parameters to be portrayed using the same map unit framework. Soil landscape mapping is a regional planning tool that identifies soil and landscape constraints to development from natural resource management and engineering viewpoints. Although soil surveys have been completed over the NSW costal regions, these are usually at scales of 1:100000 or smaller (e.g. 1:250000). This is a common fact worldwide, only a few countries with extensive land use have detailed and complete soil survey coverage (McKenzie et al. 2000). These coarser datasets are useful, to some extent, for identifying land capability on a regional basis, but these standard soil surveys were not designed to provide the fine scale resolution needed in detailed environmental planning and urban or regional development. More detailed soil landscape mapping that represents soil variation at an adequate scale (e.g. 1:25 000 or larger) is needed for orderly land use planning such as infrastructure, buildings, roads, effluent management, and other uses. In NSW, while soil landscape maps are limited by scale, soil landscape map unit descriptions in accompanying reports contain many details on soil properties and distribution patterns within the soil landscape. In many instances, internal patterns of soil variation can be ascribed to soil landscape facets.

Soil landscape facets are a way of dividing a soil landscape into discrete sub-units, each containing a distinct soil type or suite of soil types (e.g. Speight 1988; Walker 1991). Facets are homogenous and contain particular dominant soil attributes; they typically repeat in a pattern within the parent soil landscape (Chapman et al. 2004). They are delineated on the basis of their geomorphology, or other important conditions of soil formation, and their homogenous conditions are considered to be appropriate for land use planing (Speight 1988). Land system and soil landscape mappers have long recognised that the landscape may be further subdivided into discrete but unmapped landform definable facets. Facets are normally mapped at a scale of 1 : 25 000 or larger, which are too fine for 1 : 100 000 or smaller scale maps in which facets are not mapped. In NSW there are extensive areas which have land system or soil landscape mapping with textual information concerning facet distribution (Walker 1991). However, as the time and cost of soil mapping increases dramatically with finer map scales, it is often not practical to delineate facets using traditional field survey or air photo interpretation, particularly when a large mapping area is concerned. Alternative soil landscape mapping techniques are needed in order to spatially predict facets or the sub-division of soil landscape unit.

Landforms based on landscape toposequence (or topographic profile) in a catchment have been used to define landscape features to aid in soil and land capability mapping (e.g. Northcote 1978; Blaszczynski 1997; Duh and Brown 2005). They are generally classified into 4 landform classes representing shoulder (or crest), backslope (or upperslope), footslope (or lowerslope), and level (or valley) in most applications. A 4-7 landform classification schema has been proposed and implemented in FLAG (Fuzzy Landscape Analysis Geographic Information System) model (Roberts et al. 1997; Summerell et al. 2004, 2005). Further to FLAG model, a multi-resolution valley bottom flatness (MrVBF) index has been developed to identify relatively flat and low areas in the landscape at a range of scales (Gallant and Dowling 2003). Pennock et al. (1987) defined 7 distinct landform elements by measures of gradient, plan, and profile curvature. MacMillan et al. (2004) further extended the classification to 15 classes by replacing the level class with 6 separate units to differentiate level areas and depressions in upper, mid, and lower landscape positions, respectively. An extra unit was added in both the mid-slope and lower slope landscape positions that was neither markedly convex nor concave, but rather planar in the across-slope direction. Other landform classification systems (4-15 classes) have also been reported (e.g. Shary 1995; Shary et al. 2002; Carlson et al. 2004; MacMillan et al. 2004; Draguta and Blaschkeb 2006). These landforms have been proven useful in hydrologic related applications at catchment scales (e.g. Quinn et al. 1991; Moore et al. 1993; Gessler and Chadwick 2000; Summerell et al. 2004, 2005). However, the catchment-scale based landforms (based on toposequence in a catchment) often do not contain adequate details to delineate soil-landscape facets within a particular landscape unit, not are they associated with soil landscape units in soil mapping. The purpose of this study was to further delineate sub-units (facets) within a coarser soil landscape unit based on its facet occupants, rather than those landforms at a catchment scale.

This study differs in several ways from the previous studies. First, the objective of this study was to further delineate soil landscape facets (at about 1: 25 000 scale) within existing coarser soil landscape units (at a scale of 1: 100 000 or smaller), rather than a whole catchment or study area, using innovative facet division techniques and programs. Second, the delineation of soil landscape facets was guided by instructions (e.g. percentage of facet components) from field soil survey and supported by comprehensive soil databases. Third, automated GIS programs have been developed for soil landscape facet division and terrain modelling; thus, the processes are fast and repeatable, which is important when large areas (the whole coastal area in NSW in this case) are concerned. Last, the methodology was based on commonly used terrain indices (e.g. CTI) and routinely available datasets, and is applicable to other regions. The soil landscape facet division is a crucial element in the CCA project intending to provide accessible, timely, and relevant land capability information for a range of specific land uses for coastal NSW and its hinterland as defined by the mapping area. The capability information will be made available for use in urban and regional planning. Maps presenting capability or constraint information derived from soil landscape facets and associated datasets will help to identify land with high constraints or limitations so that ameliorations are taken into account in the development (Yang et al. 2005).

We initially chose the Tweed Catchment in upper northern NSW as a trial to test the methodology and implementation of the programs. The automated routines were then applied to all rural coastal catchments in the State of NSW to derive soil landscape facets and then ultimately calculate soil landscape constraint scores at the facet level. A companion paper will focus on how the land capability of each delineated soil landscape facet was assessed for multiple land uses (X. Yang, J. Gray, G. Chapman and M. Young, unpublished data). A range of constraint maps for costal NSW have been produced for various land use purposes such as standard residential, cropping, grazing, and domestic waste disposal. Currently, the methodology is being applied to inland catchments and the constraint mapping for the Hawkesbury-Nepean Catchment, a large inland catchment about 21 000 [km.sup.2], has already been completed.

Study area

The study area includes all NSW coastal catchments except the greater Sydney metropolitan area (Fig. 1). It comprises twenty 1 : 100 000 map sheets with a total area of 38074 [km.sup.2]. Most of these areas have published soil landscape maps in a scale of 1 : 100 000 or 1 : 250 000. Digital elevation model (DEM) at 25 m resolution is available for the entire coastal area.

For modelling and mapping purposes, the coastal catchments to the north of Sydney were grouped as North Coast, while those to south of Sydney were grouped as South Coast. There are 10 natural catchments in the North Coast area and 7 in the South Coast. The size of the catchments ranges from about 500 [km.sup.2] to >20 000 [km.sup.2] with an average size of 8000 [km.sup.2]. Figure 1 indicates the catchments areas and 1 : 100 000 map sheet boundaries. The actual mapped areas in North Coast and South Coast are marked in the Figure. In addition, an inland catchment, Hawkesbury-Nepean Catchment, has also been recently mapped.

Methods and procedures

The Compound Topographic Index (CTI) is a steady-state wetness index, which is also known as Topographic Wetness Index or Topographic Index, as it was initially named in Beven and Kirkby (1979). It is expressed as:

CTI = ln ([A.sub.s]/tan [beta]) (1)

where [A.sub.s] is the specific catchment area expressed as [m.sup.2] per unit width orthogonal to the flow direction, and [beta] is the slope angle expressed in radians (Beven and Kirkby 1979; Gessler et al. 1995; Wilson and Gallant 2000).

[FIGURE 1 OMITTED]

CTI is a function of both the slope and the upstream contributing area per unit width orthogonal to the flow direction. It is actually the inverse of stream-power and therefore relates to fluid flow and deposition within the landscape. We chose this index for the project because CTI has been proven to be correlated with several soil attributes such as horizon depth, silt percentage, organic matter content, and phosphorus (Moore et al. 1993). In addition, CTI is relatively easier to compute and widely used in hydrology and terrain related applications. However, the methodology and the facet division program developed in this study are not limited to CTI only, other topographic indices (e.g. FLAG UPNESS index, Roberts et al. 1997) may also be used as substitution.

The methods for delineating soil landscape facets were based on a program for prediction of performance of on-site effluent management systems described in Chapman et al. (2004). It is a batch process that first clips a CTI surface within a soil landscape unit and then examines the distribution of CTI values within that soil landscape. A cumulative frequency histogram (or cumulative distribution curve) is then constructed for each soil landscape so that the values of CTI corresponding to a specified percentage area of the landscape can be determined, a similar process to that described in Summerell et al. (2005). The area percentages have been estimated in the field and recorded in soil databases by soil surveyors from observations they have made concerning relationships between topographic position and soil type. By determining the cumulative frequency distribution of CTI across any particular soil landscape, it is possible to define CTI ranges that determine, for example, crests and other hillslope morphological types within that soil landscape. The CTI values that correspond to any particular facet are then simply given an assigned unique identifier (Facet-ID) and the relevant CTI surface is plotted as a new facet grid surface in which the facet identifier becomes the grid values (Value item in the value attribute table, VAT). The facet grid's VAT can be linked with other relevant databases through the unique identifiers or root strings. In this study, breaking up CTI into classes followed similar processes to that described previously (e.g. Summerell et al. 2005) by using the cumulative distribution curve. Additional and innovative programs have been developed to clip CTI surface based on each soil landscape unit, break facets within each soil landscape unit based on its facet occupants or percentage, and produce a seamless facet grid layer. A Visual Basic Application program has been developed as a plug-in tool in ESRI's ArcGIS environment for the above processes. The whole soil landscape facet delineating processes and, subsequently, the land constraint assessment procedures are illustrated in Fig. 2.

The dominant input datasets used in this study were 25-m DEM and soil landscape GIS data for the coastal catchments in NSW. The DEM was used to calculate slopes, flow directions, contributing areas, and CTI. Soil landscapes were used as the modelling units within which the soil landscape facets were derived based on CTI ranges. In addition, a Facet Division File (FDF) derived from soil survey databases was used in the facet division modelling process. FDF provides information on the number, percentage, and toposequence of each facet, which are needed in the facet division program for splitting each soil landscape into its component facets. A Facet Rating File (FRF), with soil landscape limiting attributes, was also prepared from soil databases and linked with GIS as external attributes to calculate the land capability or constraint scores at the facet level based on multiple soil and landscape limiting criteria (Chapman et al. 2004).

DEM with a 25-m cell size was available from the NSW Department of Lands for the whole of the NSW coastal areas. To ensure hydraulic connectivity within catchments, the DEM was processed to remove elevation anomalies (e.g. sinks and peaks) that can interfere with hydrologically correct flow (e.g. Zhou et al. 2006). In addition, null cells were filled using ArcGIS's grid focal functions (e.g. 'focalmajority') or replaced with zero for large water bodies since the facet modelling program stops at null DEM cells within the working area.

Soil landscape GIS shape files were prepared from the published 1 : 100 000 scale soil landscape maps and soil regolith data. First, the soil landscape data were topologically corrected and very small polygons or slivers (e.g. > 1 ha) were eliminated before modelling (using 'eliminate' function in ArcGIS). These slivers may cause the facet division program to crash or incorrectly attempt to divide these small polygons into facets. Then, each soil landscape polygon type was allocated a unique tag or unique root string, which was represented in the FDF and relevant soil databases. The unique root strings were created by combining map sheet numbers and soil landscape tag names so that GIS data and other relevant databases could be linked. For example, a combination of map sheet number '9232' (1 : 100 000 Newcastle Map) and soil landscape tag 'ncz' (North Arm Cove), 3 to 5 letters, creates a unique string '9232ncz'. If a soil landscape unit occurred in more than one map sheet, the map sheet number with the largest area (the dominant soil-landscape) was assigned to the new root string. This string was later used by the facet division software to group all soil landscapes of a certain type together and perform the facet division. Note that each root string must also be represented in the FDF.

[FIGURE 2 OMITTED]

For each modelling area (or catchment), a FDF was prepared from the Department's SLADE (Soil Landscape Access Database Environment) database. FDF is a comma delimited (CSV) file that contains instructions for breaking each soil landscape into its component facets. Each record in the file represents a given facet containing (1) unique soil landscape root string representing parent soil landscape, (2) facet description, (3) facet string--information about the facet's topographic position, (4) percentage of parent soil landscape occupied by this facet, and (5) unique integer identifier for this facet. The area percentages were estimated in the field by soil surveyors as a result of the observations concerning relationships between topographic position and soil type. The identifiers were assigned sequentially based on the actual number of facets in the whole study area. An example of FDF is presented in Table 1.

The contributing area ([A.sub.s] in Eqn 1) was calculated using a D-infinity (D-Inf) algorithm implemented in Terrain Analysis Using Digital Elevation Models (TauDEM) (Tarboton 1997) instead of the default D8 one as used in ESRI's ArcGIS. D8 algorithm has disadvantages arising from the proximity of flow into only 1 of 8 possible directions, separated by 45[degrees] (Fairfield and Leymarie 1991; Quinn et al. 1991; Costa-Cabral and Burges 1994). This produces unrealistic results, for example, producing striped artefacts on very gentle and long lower slopes (Gessler et al. 1995; Tarboton 1997). TauDEM calculates contributing area using single and multiple flow direction methods, overcomes the problems of loops and inconsistencies, and performs better than D8 algorithm (Tarboton 1997). After the computation of contributing area, the CTI surface for each catchment was generated using an Arc Macro Language (AML) program (CTI.AML), available free at the ESRI ArcScripts Downloads site (www.esri.com). Then, the ArcInfo 'merge' command (in Grid module) was used to produce a merged CTI grid layer for the entire North Coast catchments, and another one for the entire South Coast catchments. Gaps (or null values) along catchment boundaries or small water bodies, normally a couple of grid cells, were filled using focal functions (e.g. 'focalmajority'). The CTI surfaces generated from 25-m DEM using the D-Inf algorithm satisfactorily represent the toposequence of terrain (e.g. higher values representing drainage depressions; lower values representing the hill crests and ridges). The CTI values over the NSW North Coast range from 1.55 to 27.33 with a mean of 10.61, and 4.45 to 28.20 with a mean of 10.12 for the South Coast (Table 2).

Note that, for most coastal catchments, only up to 60% of the soil landscapes could be broken into facets (or can be modelled) based on the facet instruction from FDF. This means that there are up to 40% 'gaps', which include facets that either could not be modelled or occupy a whole soil landscapes unit. For those soil landscape units, a separate facet grid layer was prepared based on facet identifiers for those units for the north and south coasts. This grid layer was then merged with those facet grids that could be modelled (as discussed above) to form a complete facet surface.

Lastly, the merged grids were combined with the acid sulfate soil risk grid, soil erosion hazard grid, and slope gradient grid. This generated the combined grid layer containing the necessary soil landscape constraint information for various land use purposes, such as standard residential, cropping, grazing, and waste disposal. The combination was done using the 'combine' command in ArcGIS Grid module. The grids' VAT with these constraints is to be further linked with other constraint factors (e.g. mass movement, poor fertility, and flooding) to produce an overall constraint score for each 25 m by 25 m pixel and each land use category for comprehensive coastal assessment. Details of this will be presented separately (X. Yang, J. Gray, G. Chapman and M. Young, unpublished data).

Results and discussion

Soil landscape facets have been successfully generated for all of the study areas of the NSW coast. There are 1094 facets for the North Coast, 499 for the South Coast, and 539 for the Hawkesbury-Nepean Catchment. The number of facets is determined by the complexity of the parent soil landscapes and whether or not they can be modelled. Facet modelling is commonly most successful for soil landscapes with sufficiently pronounced relief for the CTI to readily discriminate terrain differences, for example, a hilly colluvial soil landscape (Alum Mountain, AMZ) as shown in Fig. 3. The black lines are soil landscape boundaries; the darkness levels represent various soil landscape facets with black representing crests (40%), grey representing upper slopes (40%), and light grey representing mid-slopes (20%). These percentages are obtained from the FDF, that is the facet division instruction from the soil surveyor. The histogram of the CTI values for this soil landscape is also presented in Fig. 3. Note that the CTI values have been rescaled to the range 0-1 for this chart. The actual CTI values for this soil landscape range from 6.74 to 20.79 with a mean of 9.23 and s.d. 1.07.

[FIGURE 3 OMITTED]

The merged CTI or facet grid often contains speckles (null cells) at the soil landscape or catchment boundaries, since the flow direction cannot be calculated for an edge cell because of uncertainty of elevation of out-catchment cells influencing flow direction. Similarly, the contributing area cannot be calculated for any cell adjacent to where the flow direction is not known because the cell with missing flow direction may flow into that grid cell. Therefore, there are always several grid cells around the edge that do not get computed. Sometimes these no-data areas extend further into the domain where flow is inwards from the edge and contributing area is unknown. This is so-called 'edge contamination' described in more detail in Tarboton (1997) and Zhou et al. (2006). In this study, edge-contaminated gaps with neighbourhoods smaller than 3 x 3, were filled using focal functions (e.g. 'focalmajority') in ArcGIS Grid module. Note that some edge-smooth programs could be used to overcome the problem, but they may cause too much generalisation and loss of detail. The role is to smooth the edges but keep the necessary details.

The accuracy of the facet map was first assessed based on broad terrain categories (e.g. crest, upperslope, lowerslope, and plain) with the use of visual assessment, peer review, and field work. The visual assessment was done, in the trial stage of the project, by plotting the facet map (with 70% transparency) on top of the hillshaded DEM and comparing the facet with the underneath terrain. This provided a means for quick assessment of facet distribution for areas of interest and it was particularly useful in the trial stage of modelling. We also compared the modelled facets with existing landform data. We generated 132 random points over the study area on the facet map where we have multi-attribute data which contain detailed landform information. We obtained the terrain attributes for each point from manual interpretation (from aerial photo or DEM) and from the terrain attributes in the multi-attribute datasets. The modelled facets are in good agreement with the multi-attribute data at the reference points as far as the broad landform categories are concerned. At the completion of the project, the accuracy of the facet mapping was reassessed against field validation. We observed 175 field points randomly at various soil landscape units and landforms and compared with the model results. The overall accuracy calculated from these observation points for the facet modelling reached 93%. However, the accuracy of facet modelling is highly dependent on the effectiveness of the facet division instruction and the soil landscape boundaries. The soil landscape boundaries may not exactly match with the terrain boundaries or the DEM may contain abnormal cells. All these are potential sources of error and need to be carefully checked and corrected.

Dividing soil landscape units into facets is not new in concept, but the implementation is rather difficult since it involves complicated spatial analysis and terrain modelling processes, particularly when a large area is concerned. This study has successfully implemented these concepts, along with innovative terrain modelling and facet delineation techniques, in a GIS environment and produced soil landscape facets in unprecedented level of details over a significantly large area across the whole costal area of NSW. The GIS programs developed in this study allow the execution of these processes to be automatic, fast, and repeatable.

The successful delineation of soil landscape facets contributed to achieving the key objectives of the soil and landscape assessment component of the CCA project. In the CCA project, the facet surfaces are further used to produce land constraint maps for a range of land uses within the coastal study area. The soil landscape constraint information, combined with other natural resource and socio-economic assessment results, will ensure the most appropriate planning decisions being made over the coastal area. A subsequent paper will present details on how the soil landscape facet information was used to assess the capability or physical potential of lands within the coastal study area for a range of land uses.

Conclusion and further studies

The study developed an appropriate methodology to produce detailed soil landscape facets from commonly available datasets over a large area. Specific programs were developed for the implementation of these methods in a GIS environment so that the process was automated, fast, and repeatable. The methodology and programs developed in this study has been proven to be efficient in delineating soil landscape facets, which allow for the identification of land constraints at levels of unprecedented detail for the coast of NSW.

It is common that some soil landscape facets are not mappable using the CTI index. In some instances, lithology changes over short distances with very little surface expression and, as such, cannot be represented using available digital elevation models. In future it is hoped that some facets will be identified by using other terrain factors such as aspect, relative elevation, solar exposure, or proximity to other features using similar methods.

We propose further studies to investigate relevant landform indices (i.e. FLAG, Roberts et al. 1997) and relief analysis models (e.g. Pennock et al. 1987; McNab 1993; Riley et al. 1999) to derive facet information for those units that were difficult to model with the current programs. We also intend to incorporate further ancillary GIS datasets and high resolution remote sensing images to aid the facet subdivision. More automated programs are also to be developed to accelerate the data preparation and analysis. After quantitative accuracy assessment and evaluation, these methods will then be applied to other inland catchments cross the whole State. Consequently, the land capability or constraint assessment and mapping will be also applied to the rest of the State.

Acknowledgments

This project was funded by the New South Wales Government and managed through the former Department of Natural Resources (DNR), now the Department of Environment and Climate Change (DECC). Many DNR staff, particularly the soil surveyors, contributed to this project and their efforts are greatly appreciated. We thank David Wainwright from WBM Pty Ltd for assistance with the facet division program and Professor David Tarboton from Utah State University for providing TauDEM program and useful comments.

Manuscript received 17 May 2007, accepted 16 October 2007

References

Beven KJ, Kirkby MJ (1979) A physically based variable contributing area model of basin hydrology. Hydrological Sciences Bulletin 24, 43-69.

Blaszczynski JS (1997) Landform characterization with Geographic Information Systems. Photogrammetric Engineering and Remote Sensing 63, 183-191.

Carlson B, Wanga D, Capena D, Thompson E (2004) An evaluation of GIS-derived landscape diversity units to guide landscape-level mapping of natural communities. Journal of Nature Conservation 12, 15-23. doi: 10.1016/j.jnc.2003.07.001

Chapman GA, Gray JM, Pavan N (2004) Using soil landscape mapping for on-site sewage risk assessment. In 'Proceedings of SuperSoil 2004 Conference, Australian and NZ Societies of Soil Science Conference'. University of Sydney, NSW, 5-9 December 2004. (CD-ROM, The Regional Institute Ltd: Sydney)

Costa-Cabral M, Burges SJ (1994) Digital Elevation Model Networks (DEMON): a model of flow over hillslopes for computation of contributing and dispersal areas. Water Resources Research 30, 1681-1692. doi: 10.1029/93WR03512

Dent D, Young A (1981) 'Soil survey and land evaluation.' (George Allen & Unwin: London)

Draguta L, Blaschkeb T (2006) Automated classification of landform elements using object-based image analysis. Geomorphology 81, 330-344. doi: 10.1016/j.geomorph.2006.04.013

Duh JD, Brown DG (2005) Generating prescribed patterns in landscape models. In 'GIS, spatial analysis and modeling'. (Eds DJ Maguire, M Batty, MF Goodchild, CA Redlands) pp. 423-444. (ESRI Press: Redland, CA)

Fairfield J, Leymarie P (1991) Drainage networks from grid digital elevation models. Water Resources Research 27, 709-717. doi: 10.1029/ 90WR02658

Gallant JC, Dowling TI (2003) A multiresolution index of valley bottom flatness for mapping depositional areas. Water Resources Research 39, 1347.

Gessler PE, Chadwick OA (2000) Modeling soil landscape and ecosystem properties using terrain attributes. Soil Science Society of America Journal 64, 2046-2056.

Gessler PE, Moore ID, McKenzie NJ, Ryan PJ (1995) Soil-landscape modeling and spatial prediction of soil attributes. International Journal of GIS 9, 421-432.

MacMillan RA, Jones KR, McNabb DH (2004) Defining a hierarchy of spatial entities for environmental analysis and modeling using digital elevation models (DEMs). Computers, Environment and Urban Systems 28, 175-200. doi: 10.1016/S0198-9715(03)00019-X

McKenzie N J, Gessler PE, Ryan PJ, O'Connell DA (2000) The role of terrain analysis in soil mapping. In 'Terrain analysis: principles and applications'. (Eds JP Wilson, JC Gallant) pp. 245-265. (John Wiley & Sons: New York)

McNab HW (1993) A topographic index to quantify the effect of mesoscale landform on site productivity. Canadian Journal of Forest Research 23, 1100-1107. doi: 10.1139/x93-140

Moore ID, Gessler PE, Nielsen GA, Peterson GA (1993) Soil attribute prediction using terrain analysis. Soil Science Society of America Journal 57, 443-452.

Northcote KH (1978) Soils and land use. In 'Atlas of Australian resources'. (Division of National Mapping: Canberra)

Pennock DJ, Zebarth BJ, De Jong E (1987) Landform classification and soil distribution in Hummocky Terrain, Saskatchewan, Canada. Geoderma 40, 297-315. doi: 10.1016/0016-7061(87)90040-1

Quinn P, Beven K, Chevallier P, Planchon O (1991) The prediction of hillslope flow paths for distributed hydrological modeling using digital terrain models. Hydrological Processes 5, 59-80. doi: 10.1002/ hyp.3360050106

Riley SJ, DeGloria SD, Elliot R (1999) A terrain ruggedness index that quantifies topographic heterogeneity. Intermountain Journal of Sciences 5, 1-4.

Roberts DW, Dowling TI, Walker J (1997) FLAG: a Fuzzy Landscape Analysis GIS Method for Dryland Salinity Assessment. CSIRO Land and Water Technical Report 8/97, Canberra.

Shary PA (1995) Land surface in gravity points classification by a complete system of curvatures. Mathematical Geology 27, 373-390. doi: 10.1007/BF02084608

Shary PA, Sharaya LS, Mitusov VA (2002) Fundamental quantitative methods of land surface analysis. Geoderma 107, 1-32. doi: 10.1016/ S0016-7061(01)00136-7

Speight JG (1988) Land classification. In 'Australian soil and land survey handbook'. (Eds RH Gunn, JA Beattie, RE Reid, RHM van de Graaf) pp. 38-57. (Inkata Press: Melbourne)

Summerell GK, Dowling TI, Wild JA, Beale G (2004) FLAG UPNESS and its application for determining seasonally wet and waterlogged soils. Australian Journal of Soil Research 42, 155-162. doi: 10.1071/SR03028

Summerell GK, Vaze J, Tuteja NK, Grayson RB, Beale G, Dowling TI (2005) Delineating the major landforms of catchments using an objective hydrological terrain analysis method. Water Resources Research 41, W12416. doi: 10.1029/2005WR004013

Tarboton DG (1997) A new method for the determination of flow directions and contributing areas in grid digital elevation models. Water Resources Research 33, 309-319. doi: 10.1029/96WR03137

Walker PJ (1991) Land systems of western NSW, Technical Report No. 25, Soil Conservation Service of NSW, Sydney, 78 pp.

Wilson JP, Gallant JC (2000) Secondary topographic attributes. In 'Terrain analysis: principles and applications'. (Eds JP Wilson, JC Gallant) pp. 87-131. (John Wiley & Sons: New York)

Yang X, Chapman G, Young M, Gray J (2005) Using compound topographic index to delineate soil facets from digital elevation models for comprehensive coastal assessment. In 'Proceedings of MODSIM05 Conference'. The University of Melbourne, Parkville, Victoria, 12-15 December 2005.

Zhou Q, Liu X, Sun Y (2006) Terrain complexity and uncertainties in grid-based digital terrain analysis. International Journal of Geographical Information Science 20, 1137-1147. doi: 10.1080/13658810600816573

X. Yang (A,B), G. A. Chapman (A), J. M. Gray (A), and M. A. Young (A)

(A) New South Wales Department of Environment and Climate Change (DECC), PO Box 3720, Parramatta, N SW 2124, Australia.

(B) Corresponding author. Email: xihua.yang@dnr.nsw.gov.au
Table l. Example of a facet dicision file (FDF)
used soil landscape facet delineation

SL_String     Description

9232bea       9232bea_upperslopes
9232bez       92326ez_crest
9232bez       9232bez_upperslope
9232bez       9232bez_hillslope
9232bez       9232bez_lowerslope
9232bez       9232bez_drainageLines
9232bfa       9232bfa_beach
9232bfz       9232bfz_flat
9232bha       9232bha_conglomerate
9232bhz       9232bhz_upperslope/crest
9232bhz       9232bhz_hillslopes
9232bhz       9232bhz_opendepression
9232bia       9232bia_crests
9232bib       9232bib_hills
9232biz       9232biz_cannot_be_modelled
9232bra       9232bra_lowerslope
9232brz       9232brz_rises
9232brz       9232brz_lowerslope
9232brz       9232brz_opendepression
9232btz       9232btz_dunefield

SL_String     Facet        Percent     Facet-ID

9232bea       9232bea1       100           1
9232bez       9232bez1        20           2
9232bez       9232bez2        20           3
9232bez       9232bez4        20           4
9232bez       9232bez5        20           5
9232bez       92326ez6        20           6
9232bfa       9232bfa1       100           7
9232bfz       9232bfz1       100           8
9232bha       92326ha1       100           9
9232bhz       92326hz1        40          10
9232bhz       92326hz3        55          11
9232bhz       9232bhz4         5          12
9232bia       92326ia1       100          13
9232bib       9232bib1       100          14
9232biz       9232biz1       100          15
9232bra       9232bra1       100          16
9232brz       9232brz1        45          17
9232brz       9232brz3        50          18
9232brz       9232brz4         5          19
9232btz       9232btz1       100          20

Table 2. Statistics of the CTI values of the NSW coast
catchments

                  Area
Name          ([km.sup.2])    Mean     s.d.    Min.    Max.

Bega              2839         9.72    1.92    5.82    26.14
Bellinger         3457        10.01    2.21    5.54    34.45
Brunswick          510        10.90    2.69    6.47    24.25
Clarence         22265         6.59    1.57    0.15    12.00
Clyde             3436         9.97    2.08    5.13    25.67
Hastings          4519        10.36    2.39    5.73    37.33
Hunter           21452         7.45    2.22    1.55    22.98
Karuah            4487        10.88    2.52    6.20    31.71
Macleay          11410        10.14    2.17    5.31    36.07
Manning           8177         9.84    2.07    5.58    33.57
Moruya            1487         9.32    1.86    5.71    26.34
Richmond          7028        11.31    2.74    5.75    33.43
Shoalhaven        7215        10.40    1.99    4.45    26.38
Towamba           2166         9.72    1.81    6.20    25.65
Tuross            2165         9.50    1.87    5.64    25.55
Tweed             1079        10.27    2.39    5.17    33.78
Gale Copyright: Copyright 2007 Gale, Cengage Learning. All rights reserved.