Disaggregation of the Central Mallee and Hopetoun Land Systems Using the Digital Elevation Model

Disaggregation of the Central Mallee and Hopetoun Land Systems using the Digital Elevation Model

April 2009

Publication details Acknowledgements Dissaggregation of the Central Mallee and The Mallee CMA and the DPI project team Hopetoun Land Systems using the Digital would like to thank retired soil scientists Jim Elevation Model Rowan and John Martin for furnishing us with their immense knowledge of the geo- Mallee Catchment ISBN No: 978-1-74217-450-1 (print) morphology and soils of the Mallee area. Management Authority ISBN No: 978-1-74217-451-8 (online) We would also like to acknowledge the www.malleecma.vic.gov.au

participation and involvement of DPI PO Box 5017 Mildura 3502 April 2009 colleagues across North–West Victoria in Telephone 03 5051 4377 Sponsor: Australian Government’s National field validating and sharing their knowledge Facsimile 03 5051 4379 Action Plan for Salinity and Water Quality on the region. Victorian Government Department of Copyright Primary Industries. The Department of Primary Industries was © Mallee Catchment Management engaged by the Mallee Catchment Authority 2008 and The State of Victoria, Authors: Jonathan Hopley, Nathan Management Authority (CMA) to undertake 2009 Robinson, Richard MacEwan and David this project with funding provided by the

Rees Australian Government’s National Action This publication is copyright. No part Department of Primary Industries Plan for Salinity and Water Quality (NAP) may be reproduced by any process Future Farming Systems Research and the Victorian Government Department except in accordance with the provisions Bendigo of Primary Industries. of the Copyright Act 1968.

Cover images Disclaimer Left: East–west dunes of the Central Mallee Publications produced by the Mallee Land System. Catchment Management Authority and Middle: Jim Rowan and John Martin with the State Government of Victoria may be DPI Staff of assistance to you but the Mallee Right: Elongated ridge in the distance Catchment Management Authority and located east of Speed. the State Government of Victoria and its Cover photos: DPI employees do not guarantee that the

publication is without flaw of any kind or is wholly appropriate for your particular purpose and therefore disclaims all liability for any error, loss or other consequence which may arise from you relying on any information in this or any Mallee Catchment Management Authority publication.

Project partners

DEPARTMENT OF PRIMARY INDUSTRIES

Contents

Acknowledgments...... iv Summary...... v 1 Introduction...... 1 1.1 Aims and objectives...... 2 1.2 The study area...... 2 2 Methodology ...... 5 2.1 Terrain analysis using rulesets on derived surfaces ...... 6 2.2 Terrain analysis using MrVBF and FLAG ...... 6 2.3 Combining the ruleset map model and the FLAG/MrVBF model outputs...... 9 2.4 Cleaning the dataset/removing noise from the dataset...... 10 2.5 Validation of model outputs...... 10 3 Results and discussion ...... 12 3.1 Disaggregation of land systems ...... 12 3.2 Descriptions of landform components ...... 16 3.3 Land system disaggregation accuracy assessment...... 27 3.4 Landform pattern regions...... 28 3.5 Area analysis and statistics of the landform components ...... 29 4 Conclusion ...... 32 4.1 Recommendations...... 32 References ...... 34 Appendix 1 Field validation route map...... 35

Figures Figure 1 The Central Mallee and Hopetoun land systems highlighted from Rowan and Downes (1963) mapping...... 2 Figure 2 Central Mallee land system terrain section (from Rowan and Downes 1963) ...... 3 Figure 3 Hopetoun land system terrain section (from Rowan and Downes 1963) ...... 4 Figure 4 A chart summarising the land disaggregation methodology...... 6 Figure 5 Mosaic of images showing various surface layers derived from the central DEM elevation dataset. The same view, presented in each image, is of a linear east–west dune in the Central Mallee Land System ...... 7 Figure 6 Sectors of the Central Mallee and Hopetoun land systems where the MrVBF and FLAG models were trialled...... 7 Figure 7 UPNESS index cdf log plot with three points discriminating four landform elements (source: Summerell et al. 2005)...... 9

i DISAGGREGATION OF MALLEE LAND SYSTEMS

Figure 8 A section of the Central Mallee Land System showing east–west dunes before cleaning (left) and after cleaning (right). Notice how the missing cells of the crests have been filled and the small areas previously assigned as dunes have been removed ...... 10 Figure 9 Sectors of analysis for ridge and lakebed landform components. Each sector has a slightly different elevation threshold applied by the ruleset...... 12 Figure 10 Example area of the Hopetoun Land System mapping with the disaggregated landform components. All landform components are represented in this map ...... 14 Figure 11 East–west dunes in the Central Mallee Land System. The left image shows the classified dunes (yellow border) superimposed over the relative elevation (70 m) surface, the right image shows the same dunes superimposed over the aerial photography. The north–south extents of the dunes seem to match well with the imagery; however the areas identified in the red boxes show an under‐ classification in this area...... 15 Figure 12 Extent of the classified ridges in a part of the Hopetoun Land System superimposed over a hill‐ shading of the DEM. The red box on the left highlights ridge areas that have not been classified as ridges whilst the box on the right shows a large lunette that was later classified as ‘lunettes associated with ridge’...... 15 Figure 13 East–west dunes of the Central Mallee Land System...... 17 Figure 14 East–west dunes on the western slopes of a ridge...... 18 Figure 15 Elongated ridge in the distance located east of Speed...... 19 Figure 16 Gentle rise located on an elongated ridge...... 20 Figure 17 Lower ridge slopes located east of Speed...... 21 Figure 18 Gentle rise on an undulating plain east of Ouyen ...... 22 Figure 19 Interdune swale (corridor) near Manangatang ...... 23 Figure 20 Lunettes associated with groundwater depressions near Murrayville...... 24 Figure 21 Lunette associated with ridge (yellow dashed line), prominent former lakebed (red dashed line) and ridge (aqua dashed line) near Woomelang ...... 25 Figure 22 Former lakebed located east of Rainbow ...... 26 Figure 23 Percentage cover of the east–west linear dunes ...... 30 Figure 24 Map showing the landform pattern regions and their area as described in Table 5...... 30

Tables

Table 1 Some metadata relating to the project’s core data input, a 10m digital elevation model...... 5 Table 2 Classification of FLAG/MrVBF categories into revised classes ...... 9 Table 3 Rules used to disaggregate landform components in the Central Mallee and Hopetoun land systems...... 13 Table 4 Final Central Mallee and Hopetoun landform component classes ...... 16 Table 5 Validation results for each landform class. Incorrect samples were assigned by visual assessment to the next most likely class ...... 28 Table 6 Landform pattern regions in the Central Mallee (CM) and Hopetoun (H) land systems...... 29 Table 7 Area cover (expressed in percentage) of each mapped landform component in each landform pattern region (mapped in Figure 24)...... 31 DEPARTMENT OF PRIMARY INDUSTRIES

iii DISAGGREGATION OF MALLEE LAND SYSTEMS

Acknowledgments The project team would like to thank retired soil scientists Jim Rowan and John Martin for furnishing us with their immense knowledge of the geomorphology and soils of the study area. We would like to acknowledge the participation and involvement of DPI colleagues across north‐ western Victoria in field validating and sharing their knowledge on the region. The contributions of Leisa Macartney and Colin Smith in reviewing this report are also greatly appreciated. The authors would also like to acknowledge the support of the Mallee Catchment Management Authority and the National Action Plan for Salinity and Water Quality, and the Department of Primary Industries.

iv DEPARTMENT OF PRIMARY INDUSTRIES

Summary The description of Mallee land systems provided by Rowan and Downes (1963) still serves as the most used terrain interpretation of the Victorian Mallee. Whilst providing broad‐scale information about landform components (dimensions and distribution), the land systems at a nominal scale of 1:250,000 don’t spatially identify individual landforms in the landscape. The dissection of land systems into landform components (patterns or elements) utilising secondary derivatives such as slope and aspect from a DEM has been successful for the Tempy Land System of the Mallee (MacEwan et al. 2007). This project seeks to further develop this methodology by incorporating the FLAG (Fuzzy Landscape Analysis GIS) model and MrVBF (Multi‐resolution Valley Bottom Floor) index and applying these to the Central Mallee and Hopetoun land systems. The delineation of landform components for the two land systems was undertaken using a combination of DEM applied rules and FLAG and MrVBF model outputs for sectors where quality landform mapping was identified. A statistical validation approach was undertaken to assess quality of model outputs and further refine mapping to improve landform component predictions. Combining each of the model outputs through a cognitive selection has provided an enhanced final output that contains the desired elements from each model. Analysis of the map indicates areas where landform components have been over‐ and under‐classified. An overall accuracy of 74% was recorded with 7 of the 10 land components (comprising 97% of the study area) achieving an 84% accuracy, while the remaining three land components achieved an accuracy of 50%. These three landform components were reviewed in light of the field validation and were anticipated to have significantly improved the overall map accuracy. The combined approach of using DEM applied rules and model outputs from the FLAG and MrVBF have proven reliable and detailed in the delineation of landform components for the Central Mallee and Hopetoun land systems. Completion of land system disaggregation to land components for the remaining land systems of the Mallee will provide landform data at a resolution at which effective modelling and assessment of land management issues—such as susceptibility to wind erosion—can occur.

v DISAGGREGATION OF MALLEE LAND SYSTEMS

Disaggregation of the Central Mallee and Hopetoun land systems using the digital elevation model

Jonathan Hopley, Nathan Robinson, Richard MacEwan, David Rees

1 Introduction

The Mallee region has a diversity of soils and landforms that support a range of primary production enterprises and provide refuges, reserves and parks for native flora and fauna. Management of natural resources in the Mallee region requires an understanding of the processes that have shaped this landscape and that affect its productivity and its vulnerability to degradation. The work carried out in this project contributes to that understanding and builds on the foundation soil and land survey work carried out in the 1950s (Rowan & Downes 1963) as well as subsequent revisions to land systems mapping by Jim Rowan in the 1980’s and the recent development of the Victorian Geomorphology Framework. This report describes the methodology developed to disaggregate the Central Mallee and Hopetoun land systems of Rowan and Downes (1963) into their landform components and presents the map outputs derived from this method. The report also makes some recommendations for future work based on the methodology and utilisation of the data outputs. An increase in the detail of soil and landform mapping would provide an excellent data input to: enhance the accuracy and precision of modelling; monitor land degradation with greater certainty; conduct agronomic trials on representative soils and landscapes; and further prioritise extension and research activities to deal with land use/land erosion susceptibility conflict. The project was funded by the Mallee Catchment Management Authority (MCMA) and Department of Primary Industries (DPI) project ‘Capturing soil landscape parameters that underpin regional biophysical modelling’ and was undertaken between September and December 2008. To date, the description of Mallee land systems provided by Rowan and Downes (1963) still serves as the most used terrain interpretation of the Victorian Mallee. These descriptions or idealised sections provide valuable interpretations of landform components that occur within each land system. Approximate percentage of landform components occurring within each land system is also described. While soil distribution across the aeolian landscapes of the Victorian Mallee is extremely diverse and mixed, the landform components within these idealised sections serve as a basis for defining ‘likely’ soil occurrence and associated land degradation and productive capability. Whilst providing broad‐scale information about landform components (dimensions and distribution), the land systems at a nominal scale of 1:250,000 do not provide the spatial detail required to identify individual landforms in the landscape. This effectively reduces the resolution at which effective modelling and assessment of land management issues—such as susceptibility to wind erosion—can occur. The recent acquisition and utilisation of improved land cover information derived from remote sensed Landsat data has further highlighted the need for a complimentary high resolution landform dataset to be created (MacEwan et al. 2007). Terrain of the Victorian Mallee region is relatively subdued in amplitude with subtle variations with respect to geological interactions in comparison to other parts of Victoria (such as the Western and Eastern Uplands). Therefore a digital elevation model (DEM) with sufficient detail in resolution and accuracy (vertical and horizontal) is required to distinguish landform components across the Mallee land systems. The availability of relatively new aerial photography and an orthophoto‐derived DEM at a spatial resolution of 10 m presents an opportunity to disaggregate land system units into land components using a geographical information system (GIS).

1 DEPARTMENT OF PRIMARY INDUSTRIES

The dissection of land systems into landform components (patterns or elements) utilising secondary derivatives such as slope and aspect from a DEM has been successfully trialled for the Tempy Land System of the Mallee (MacEwan et al. 2007). This project seeks to further develop this methodology by incorporating the FLAG (Fuzzy Landscape Analysis GIS) model (Roberts, Dowling & Walker 1997) and MrVBF (Multi‐resolution Valley Bottom Floor) index (Gallant & Dowling 2003) and applying these to the Central Mallee and Hopetoun land systems. These models were trialled in the Corangamite Catchment Management Authority (CMA) region (Robinson 2008) to map landform components and assign waterlogging susceptibility ratings to these landforms from 1:100,000 soil‐landform mapping.

1.1 Aims and objectives

The aim of this project is to: • create a spatial dataset comprising the landform components contained within the two land systems as described by Rowan and Downes (1963) • test the utility of the FLAG model and MrVBF index in dissecting the land systems, and if possible incorporate them into the method previously developed by MacEwan et al. (2007) • identify and delineate regions within each land system that contain similar landform patterns (landform components and their distribution).

1.2 The study area

The study area (Figure 1) is defined as the extent of the Central Mallee and Hopetoun land systems, as originally defined by Rowan and Downes (1963). This area was identified by the MCMA and represents approximately 40% of the total area of the Mallee catchment.

Figure 1 The Central Mallee and Hopetoun land systems highlighted from Rowan and Downes (1963) mapping

2 DISAGGREGATION OF MALLEE LAND SYSTEMS

The Central Mallee land system The Central Mallee Land System (Figure 2) covers an area of approximately 11,500 km2 extending from the South Australian border in the west to the River Murray in the east. Towards the west it is divided into two zones separated by the Berrook Land System. The landscape is dominated by east–west trending dunes which occur at a greater density here than in any other cleared parts of the Mallee. The dunes occupy on average 30% of the landscape, which corresponds to three or four dunes in a north‐south transect of 1 km, but their density varies across the land system. The basement on which the dunes rest is level to undulating plains, hummocks and NNW–SSE trending ridges. The dunes have an average relief of 5 m and the ridges 15 m (although these too vary across the land system). The Central Mallee Land System has a high erosion hazard due mainly to its dense array of east–west dunes. The hazard is also severe on the sands of the upper slopes of hummocks and ridges where the western facing slopes are stripped of surface soil and the eastern faces are areas of drift accumulation. The predominant form of erosion is aeolian.

Figure 2 Central Mallee land system terrain section (from Rowan and Downes 1963)

The Hopetoun land system The Hopetoun Land System (Figure 3) covers an area of approximately 4,000 km2. The bulk of the land system occurs as a unit extending from Rainbow to Sea Lake. It also occurs in scattered areas further northwards and westwards, particularly to the north of Lake Tyrrell, to the south‐west of Murrayville and to the north‐west of Ouyen. The predominant landscape is composed of a regular series of NNW–SSE trending ridges on which dunes are superimposed, and of inter‐ridge plains (Figure 3). The ridges occur at a similar density to those in the Tempy Land System, the average distance between their crests being about 3 kmand they are significantly more pronounced than those that persist in the Central Mallee with their relief being up to 25 m. The dunes are weakly developed and relatively few, occupying only about 5% of the landscape compared with 15% in the Tempy Land System and 30% in the Central Mallee Land System. The dunes are almost completely absent in the southern reaches of the land system. The overall erosion hazard within the land system is considerably less than that in the Central Mallee and Tempy land systems, but greater than that in the Culgoa Land System (Rowan and Downes 1963). The

3 DEPARTMENT OF PRIMARY INDUSTRIES

hazard is greatest on the dunes and generally moderate for the sandy loams of the ridges, except on the upper westerly‐facing slopes which are most susceptible to wind stripping. While lakebeds and lunettes do not feature in Rowan and Downes’ (1963) earlier descriptions of these two land systems, the development of the map model together with availability of higher resolution satellite and aerial imagery have allowed these to be identified. For this reason they have also included in the land system disaggregation.

Figure 3 Hopetoun land system terrain section (from Rowan and Downes 1963)

4 DISAGGREGATION OF MALLEE LAND SYSTEMS

2 Methodology

The disaggregation of the land systems into landform components was undertaken through terrain analysis techniques using a digital elevation model (DEM). The DEM was provided by the Department of Sustainability and Environment (DSE), and the metadata is shown in Table 1.

Table 1 Some metadata relating to the project’s core data input, a 10m digital elevation model.

Dataset name: Vicmap Elevation DEM 10 m Custodian: Department of Sustainability and Environment Grid cell size: 10 m Currency: 30 June 2008 Positional accuracy: 12.5 m horizontal, +/−5 m vertical or better

The DEM was found to be of a poorer quality in the public land area contained in the western section of the Central Mallee Land System (the Murray Sunset National Park) and as such this area has been excluded from the study. This is possibly due to the lower quality elevation data acquired for this area which appears to be sourced from the Shuttle Radar Digital Terrain Model (refer to http://srtm.csi.cgiar.org/ for further information). The analysis involved the following key steps which are summarised in Figure 4, and explained in more detail below in the following sections: 1. Clipping the DEM to Rowan and Downes’ (1963) mapping for the Central Mallee and Hopetoun land systems. 2. Generating derivative datasets from the clipped DEM including slope, aspect, curvature and relative elevation. 3. Developing and applying rulesets (value thresholds) to combine the DEM derivatives to present position in the landscape of the component or critical dimensions that define it. 4. Using these preliminary outputs, dividing the land systems into sectors representing similar baseline elevation and then refining rulesets in order to achieve best outcomes across the study area. 5. Ground‐check model outputs for distribution, accuracy and inconsistencies. 6. Applying MrVBF index and FLAG model to sections of the DEM (limitations in dataset size, data connectivity and the appropriateness of applying these tools meant that selected sections of the land systems were chosen). 7. Combine MrVBF, FLAG and ruleset outputs to produce a single landform raster dataset. 8. Refine rules as necessary and combine to improve visual assessment of modelled landform components. 9. Cleaning the dataset to remove background noise (spatial filtering process). 10. Delineate regions of similar landform patterns. The selection of rule thresholds and the combination of FLAG and rule outputs was guided by: visual assessment of the DEM; other datasets such as aerial and satellite images; Rowan and Downes’ (1963) descriptions of the landforms; field visits and expert opinion.

5 DEPARTMENT OF PRIMARY INDUSTRIES

Figure 4 A chart summarising the land disaggregation methodology

2.1 Terrain analysis using rulesets on derived surfaces

Using GIS, a number of surface derivatives were created from the input DEM. These included slope, aspect, curvature and relative elevation (with a variety of search radii). See Figure 5 below for an example of each surface type.

Slope Calculated from the maximum rate of change between each grid cell and its neighbours; expressed as a percentage.

Curvature Curvature is a second derivative of the terrain surface calculated from the curvature of a surface at each cell centres. A positive curvature indicates the surface is upwardly convex at that cell relative to neighbouring cells. A negative curvature indicates the surface is upwardly concave at that cell.

Aspect Identifies the downslope direction of the maximum rate of change in value from each cell to its neighbours. Aspect can be thought of as the slope direction.

Relative elevation Calculated by comparing the value of the focus cell to the neighbouring cells to determine if it is relatively low or elevated. The range of the neighbourhood (the focal range) can be set at different radii. In this project a circular range was trialled at 5, 7, 10, 15, 50 and 100 cells (50, 70, 100, 150, 500 and 1000 metres respectively). The derivative surfaces were explored to determine their relation to the land components described by Rowan and Downes (1963). Simple algebra based on value thresholds (rulesets) was then used to combine elements from relevant surfaces to map out the components. This was an iterative process with the DEM, aerial photographs and Landsat data together with field visits, expert opinion and documentation guiding the refinement of the rules.

2.2 Terrain analysis using MrVBF and FLAG

To further enhance the outputs from the ruleset model described above, the MrVBF index and the FLAG model were trialled over selected areas of the land systems (Figure 6). These areas were chosen as they were contiguous (a constraint of FLAG) and they occurred in areas where there was relatively high variation in relief (i.e. where ridges and inter‐ridge features occurred).

6 DISAGGREGATION OF MALLEE LAND SYSTEMS

Figure 5 Mosaic of images showing various surface layers derived from the central DEM elevation dataset. The same view, presented in each image, is of a linear east–west dune in the Central Mallee Land System

Figure 6 Sectors of the Central Mallee and Hopetoun land systems where the MrVBF and FLAG models were trialled

Multi resolution Valley Bottom Flatness (MrVBF) index The MrVBF index (Gallant & Dowling 2003) specifically defines and distinguishes valley bottoms from hillslopes at a range of scales and combines landscape values into a single index. The method uses DEM slope for hydrological convergent areas and progressive DEM deterioration/generalisation procedures in reduction of slope thresholds, enabling delineation of valley forms. A number of processing iterations are undertaken to develop a measure of the valley floor extent at different scales (resolution) to allow broad‐

7 DEPARTMENT OF PRIMARY INDUSTRIES

scale valley bottom flatness to override finer scale features of unnecessary detail. These results at different scales are combined into a single index

Fuzzy Landscape Analysis GIS (FLAG) UPNESS index The FLAG model is especially useful in landscape delineation and identifying position in the landscape relative to other points in the terrain. The fuzzy modelling approach avoids scaling issues and enables specific cut‐off values to characterise the landscape (Roberts, Dowling & Walker 1997). Fuzzy set theory assigns values of membership to each grid cell on a 0 to 1 scale, 1 representing highest membership while 0 equates to lowest or poorest membership. The UPNESS index (described in further detail by Roberts, Dowling & Walker 1997; Summerell et al. 2004, 2005; Murphy et al. 2005) is a fuzzy set defined from the ‘fraction of the total landscape monotonically uphill from each pixel’. The UPNESS index values are used to discriminate landform types of the landscape based upon concave and convex break‐of‐slope inflection points. The probability distribution function (pdf) allows ready identification of inflection points (break/significant change of slope) and disaggregation of toposequences into three regions representing four different landform components. The cumulative distribution function (cdf) is used in combination to define maximum (ridge/crest tops) and minimum (valley floor) points of the sequence. A mid point between inflection points is assigned to differentiate between upper slopes and lower slopes within a toposequence (Figure 7).

Integrating FLAG and MrVBF Integration of the terrain model applications (MrVBF and FLAG) as described by Murphy et al. (2005) provides ‘an overall better landform delineation procedure’ capturing the strengths of both models. Here the MrVBF index is especially useful to map depositional areas within the landscape focussing on valley floors at multiple scales, while the FLAG landforms derived from the UPNESS index attempts to represent landforms associated with hillslopes. The integration of the two terrain model applications assigns seven landform categories (LF): • ridge tops (LF1) • upper slopes (LF2) • mid slopes (LF3) • lower slopes (LF4) • colluvial valley fill in upland landscapes or depressions (LF5) • rises in low slope alluvial fill or long gentle sloping foot slopes (LF6) • large expanses of in‐filled valleys and alluvial depositions (LF7). The FLAG landforms delineates landforms LF2 to LF4 while landforms LF5 to LF7 are defined using the MrVBF to define these depositional areas. LF1 is defined using the MrVBF as this application will separate the ridge tops from the upper slopes where the FLAG landforms cannot.

8 DISAGGREGATION OF MALLEE LAND SYSTEMS

Figure 7 UPNESS index cdf log plot with three points discriminating four landform elements (source: Summerell et al. 2005).

2.3 Combining the ruleset map model and the FLAG/MrVBF model outputs

An assessment on each of the model outputs against the aerial photography and DEM showed that each modelling technique had particular strengths in different parts of the landscape. The combined FLAG/MrVBF outputs were found to provide an increased level of landform definition in the Hopetoun areas (sectors 1 and 2) where the ridges were more prominent and the overall terrain relief greater, however they did not perform as well in the more gently undulating terrain of the Central Mallee (sectors 3 and 4). The particular strength of FLAG/MrVBF across the study area was in delineating the different landform components of the ridges (ridge top; upper, mid and lower slopes), therefore producing the most accurate map outputs for the FLAG sectors (1–4).

Flag Sectors 1 and 2 • The FLAG/MrVBF outputs provided the base layer. • The FLAG/MrVBF landform classifications were reclassified into the categories shown in Table 2. • The extent of the prominent ridge class was refined using the ruleset output for ‘prominent ridges’ and a visual assessment of the DEM. Classified ridges outside this extent were reclassed to ‘undulating plains’ • The east–west dunes, the rises on ridges and the lakebed/depression ruleset outputs were then overlaid. Table 2 Classification of FLAG/MrVBF categories into revised classes

FLAG/MrVBF landform (LF) classification Revised landform classification LF1, LF2 and LF3 Prominent ridges LF4 Lower ridge slopes LF5, LF6 and LF7 Undulating plains Flag Sector 3 • The ruleset outputs provided the base layer.

9 DEPARTMENT OF PRIMARY INDUSTRIES

• Where the prominent north–south rises occur in this sector the FLAG classes LF1–LF3 were combined and overlaid. These rises were delineated by hand based on the DEM.

Flag Sector 4 • The ruleset outputs provided the base layer. • The extent of the ridge class was refined using the reclassified ridges (LF1–LF3) of FLAG/MrVBF and a visual assessment of the DEM guided the extent

Remaining area of the study extent • The ruleset outputs were used with no further refinement. • Identified rises that were associated with lakebeds were manually defined as lunettes as were other rises that could be discerned by a visual assessment of the DEM and its derivatives.

2.4 Cleaning the dataset/removing noise from the dataset

The derived datasets representing each land component contained an amount of background noise (e.g. missing cells). This noise resulted from small areas of a land component not conforming to the rulesets used to define them or, conversely, from other cells in the DEM containing the same properties as a defined component without actually being a member of that component (incorrectly classified). Using the landform descriptions together with visual analysis of the DEM and the aerial photography, area thresholds for cleaning the data were set in order to remove this noise. The thresholds ranged from 300 to 500 connected cells (i.e. an area of up to 250 m by 250 m). ArcScan was then used to clean each dataset by filling and removing cells less than the specified threshold. These generalisation and cleaning methods are not perfect and result in some cells being incorrectly removed and others left, however overall a large amount of the background noise was successfully removed and landform components better defined, (e.g. gaps in dune tops filled, see Figure 8). A focal majority algorithm was then run on each dataset to further smooth boundaries and to fill any non‐classified cells from the previous processing.

2.5 Validation of model outputs

A field validation exercise was undertaken to measure the accuracy of the model outputs. The validation methodology involved a stratified sampling approach with sample points being randomly selected within a 50 m radius of a defined route traversing the Mallee. Thirty sample points, each representing a 10 m cell of

10 DISAGGREGATION OF MALLEE LAND SYSTEMS the original DEM, were randomly generated for each landform component. Appendix 1 shows a map of the field validation route and location of the 300 stratified random sample points. Each sample point was visually assessed for its likely membership to the identified landform class. Both the location of the sample point and the context of the surrounding landscape were used to inform the model validation assessment. If a sample point did not belong to the classified landform class then the most likely landform class was assigned.

11 DEPARTMENT OF PRIMARY INDUSTRIES

3 Results and discussion

3.1 Disaggregation of land systems

From the disaggregation modelling process, it was apparent that the rule model outputs tended to over‐ or under‐classify landform components that were elevation dependent, for example former lakebeds and ridges. To account for this, the land systems were divided into analysis sectors (Figure 9) for those particular components (refer to Table 2). In this study, relative elevation (70 m) was found to be one of the most useful derivatives for delineating landform components; whilst curvature was not used in generation of any of the final rulesets. The subtle nature of the target landform components meant that slope and curvature thresholds had to be set low in order to account for their shape. This, however, led to gross over‐classification with other non‐descript features being included in the outputs. Slope was found to be useful in identifying the former lakebeds where a low threshold was combined with a low elevation to differentiate these depressed flat areas from the surrounding undulating plains with hummocks and other finer‐scale rises. Overall, the linear dunes and prominent lakebed depressions were well mapped using this technique. The ridges were under‐ and over‐classified across the land systems with mis‐classifications arising from their changing relief and undulating slopes. The outputs derived from these rulesets were further refined through comparison and combination with MrVBF and FLAG outputs. Table 3 provides the final rulesets used to delineate each landform component. Prominent lunettes were defined within the ‘rises on undulating plains’ (RUP) landform. The lunette landform components were reduced in extent due to the relative location to the prominent former lakebeds and their crescent shape. Lunettes associated with ridges (LAR) were delineated within the prominent ridges landform as these ridges often have crescent shaped lunettes superimposed on their western slopes.

Figure 9 Sectors of analysis for ridge and lakebed landform components. Each sector has a slightly different elevation threshold applied by the ruleset

12 DISAGGREGATION OF MALLEE LAND SYSTEMS

Table 3 Rules used to disaggregate landform components in the Central Mallee and Hopetoun land systems

Landform component Sector Derived rules E‐W dunes on plains All Relative elevation (70 m) > 0.6 and aspect = between 315 and 45 or 135–225 E‐W dunes on ridges All E‐W dunes clipped to ridge extent Prominent ridges 1 Elevation > 75 m 2 Elevation > 80 m 3 Elevation > 95 m 4 Elevation > 87 m 5 Elevation > 90 m 6 Elevation > 85 m Rises on ridges 1 Elevation > 75 m and relative elevation (150 m) > 0.7 2 Elevation > 80 m and relative elevation (150 m) > 0.7 3 Elevation > 95 m and relative elevation (150 m) > 0.7 4 Elevation > 87 m and relative elevation (150 m) > 0.7 5 Elevation > 90 m and relative elevation (150 m) > 0.7 6 Elevation > 85 m and relative elevation (150 m) > 0.7 Undulating plains 1 Elevation < 75 m 2 Elevation < 80 m 3 Elevation < 95 m 4 Elevation < 87 m 5 Elevation < 90 m 6 Elevation < 85 m Rises on undulating plains 1 Elevation < 75 m and relative elevation (70 m) > 0.5 2 Elevation < 75 m and relative elevation (70 m) > 0.5 3 Elevation < 80 m and relative elevation (70 m) > 0.5 4 Elevation < 95 m and relative elevation (70 m) > 0.5 5 Elevation < 87 m and relative elevation (70 m) > 0.5 6 Elevation < 90 m and relative elevation (70 m) > 0.5 Prominent former lakebeds 1 Elevation < 45 m and slope < 2% 2 Elevation < 50 m and slope < 2% 3 Elevation < 58 m and slope < 2% 4 Elevation < 40 m and slope < 2% 5 Elevation < 54 m and slope < 2% 6 Elevation < 42 m and slope < 2% 7 Elevation < 63 m and slope < 2% 8 Elevation < 52 m and slope < 2% 9 Elevation < 53.5 m and slope < 2% 10 Elevation < 51 m and slope < 2% 11 Elevation < 50 m and slope < 2% 12 Elevation < 60 m and slope < 2% 13 Elevation < 70 m and slope < 2% 14 Elevation < 76 m and slope < 2%

The final output is a raster dataset derived from merging each of the individual land component grids. All cells within the initial starting DEM grid have been allocated to one of the landform components described in Section 3.2. A map displaying the land components, land system extents and the landform pattern regions has been created and is supplementary to this report. An area of the Hopetoun Land System containing of the landform component categories has been magnified below (Figure 10).

13 DEPARTMENT OF PRIMARY INDUSTRIES

Figure 10 Example area of the Hopetoun Land System mapping with the disaggregated landform components. All landform components are represented in this map Each of the modelling techniques employed—the rule‐based and the FLAG/MrVBF models—had strengths in delineating different landform components. The rulesets disaggregated the linear dunes, prominent former lakebeds and undulating plains well, whilst the FLAG/MrVBF provided better quality definition of the ridges and the inter‐ridge features, especially where significant terrain relief occurred. Combining each of the model outputs through a cognitive selection has provided an enhanced final output that contains the desired elements from each model. Some analysis of the map indicates areas where landform components have been over‐ and under‐ classified. This is inevitable due to the small scale variations in the topography of the landscape and the large extent to which the disaggregation has been applied. An example of this is provided in where dunes have been unde‐classified. In other areas there is an over‐classification of dunes which are likely to be hummocks or slopes of remnant ridges. Generally the linear dune extents match extremely well to the sandy upper slopes and tops of the dunes. Other small errors in the dataset also occur due to artefacts including orthophoto tile mosaic lines in the DEM. The presence of artificial elements in the landscape such as roads, towns, dams and other infrastructure can also lead to misclassification of the raster grid cells. Considering the size of the land systems in relation to these elements it is suggested these small artefacts can largely be ignored.

14 DISAGGREGATION OF MALLEE LAND SYSTEMS

Figure 11 East–west dunes in the Central Mallee Land System. The left image shows the classified dunes (yellow border) superimposed over the relative elevation (70 m) surface, the right image shows the same dunes superimposed over the aerial photography. The north–south extents of the dunes seem to match well with the imagery; however the areas identified in the red boxes show an under‐classification in this area

Figure 12 Extent of the classified ridges in a part of the Hopetoun Land System superimposed over a hill‐shading of the DEM. The red box on the left highlights ridge areas that have not been classified as ridges whilst the box on the right shows a large lunette that was later classified as ‘lunettes associated with ridge’.

15 DEPARTMENT OF PRIMARY INDUSTRIES

3.2 Descriptions of landform components

The morphology and variation of the landforms disaggregated in the Central Mallee and Hopetoun land systems as listed in Table 4. These are largely derived from descriptions provided in Rowan and Downes (1963). Table 4 Final Central Mallee and Hopetoun landform component classes

Landform component class Landform description 1 East–west dunes on plains 2 East–west dunes on ridges 3 Prominent ridges 4 Rises on ridges (these are areas on ridges that are relatively elevated and represent ridge tops and hummocks on ridge slopes) 5 Lower ridge slopes 6 Rises on undulating plains (inc: hummocks, lunettes and other rises) 7 Undulating plains 8 Prominent lunettes 9 Lunettes associated with ridges 10 Prominent former lakebeds

16 DISAGGREGATION OF MALLEE LAND SYSTEMS

East–west dunes on plains (component class 1) The dunes generally range from 80 to 120 m wide and vary in length and dune density (Figure 13). Dune relief is often 5 to 10 m with many having flat crests due to cultivation and wind erosion. Dunes have been defined using relative elevation derivatives from the DEM.

Figure 13 East–west dunes of the Central Mallee Land System

17 DEPARTMENT OF PRIMARY INDUSTRIES

East–west dunes on ridges (component class 2) Theses dunes have been identified as occurring predominantly on the western flanks of ridges (Figure 14) that often terminate at the ridge crest.

Figure 14 East–west dunes on the western slopes of a ridge

18 DISAGGREGATION OF MALLEE LAND SYSTEMS

Prominent ridges (component class 3) The ridges are elongated linear features (Figure 15) with a NNW–SSE orientation that range in breadth from 500 to 3000 m across. Often the ridges occur as a complex unit with a relatively flat crest (elevated plateaux) with a diversity of heavy and light textured soils. In the Central Mallee Land System, ridges generally are of a lower density than in the Hopetoun Land System; however this may be due to extensive deposits of the Woorinen Formation blanketing the current landscape. Ridges in the Hopetoun Land System exhibit greater vertical relief and are longer than those of the Central Mallee Land System.

Figure 15 Elongated ridge in the distance located east of Speed

19 DEPARTMENT OF PRIMARY INDUSTRIES

Rises on ridges (component class 4) These rises (Figure 16) occur on relatively flat ridge crests as well as sharp ridge crests and hummocks on easterly‐ and westerly‐facing ridge slopes.

Figure 16 Gentle rise located on an elongated ridge

20 DISAGGREGATION OF MALLEE LAND SYSTEMS

Lower ridge slopes (component class 5) The lower ridge slopes (Figure 17) were identified and mapped using the modelled lower slope output (LF4) from the FLAG model. These gentle slopes are generally < 2% and grade into rises and plains between and on ridges.

Figure 17 Lower ridge slopes located east of Speed

21 DEPARTMENT OF PRIMARY INDUSTRIES

Rises on undulating plains (component class 6) Rises found on the undulating plains (Figure 18) include hummocks, lunettes, ridge remnants and other gentle landscape undulations. The occurrence of these rises on the plains varies from area to area along with the relief and soils that occur here. An example is the steep ridge‐like landforms that occur in the western extents of the Central Mallee Land System compared with the gentle broad rises found in the east of the land system. Also identified within this landform are areas within the undulating plains that occur relatively higher than the neighbouring plains, including elevated surrounding plains of depressions.

Figure 18 Gentle rise on an undulating plain east of Ouyen

22 DISAGGREGATION OF MALLEE LAND SYSTEMS

Undulating plains (component class 7) These plains include inter‐dune corridors (Figure 19), former lakebeds and gentle slopes grading from rises (including hummocks) that occur throughout the Central Mallee and Hopetoun land systems. These occupy the lowest part of the landscapes with current copi islands and depressions typical of the Raak Land System. Plains were defined from low relative elevation values (< 0.5).

Figure 19 Interdune swale (corridor) near Manangatang

23 DEPARTMENT OF PRIMARY INDUSTRIES

Prominent lunettes (component class 8) These prominent rises associated with mapped lakes (generally occurring on the eastern side) have been assigned to this class. The lunettes (Figure 20) are generally sickle‐shaped and vary in breadth and height, often occurring as a sequence of lunettes formed from variations in palaeo‐climate. Mapping has identified numerous lunettes in the east of the Central Mallee Land System.

Figure 20 Lunettes associated with groundwater depressions near Murrayville

24 DISAGGREGATION OF MALLEE LAND SYSTEMS

Lunettes associated with ridges (component class 9) These lunettes have been identified as part of a ridge landscape complex and occur with a prominent former lakebed to the west. The crescent shaped lunettes allow this feature to be identified in parallel with an abutting ridge (Figure 21). This landform feature has largely been manually delineated through visual observation and assessment of the DEM. Further soil and geological analysis would be required to separate the ridge and lunette landforms.

Figure 21 Lunette associated with ridge (yellow dashed line), prominent former lakebed (red dashed line) and ridge (aqua dashed line) near Woomelang

25 DEPARTMENT OF PRIMARY INDUSTRIES

Prominent former lakebeds (component class 10) Flat, depressed and often circular areas have been mapped as prominent former lakebeds. This landform feature has potentially been under‐classified as it is recognised that many former lakebeds occur across the landscape. The flat areas are often associated with lunettes and are found adjacent to groundwater depressions of the Raak Land System (Figure 22).

Figure 22 Former lakebed located east of Rainbow

26 DISAGGREGATION OF MALLEE LAND SYSTEMS

3.3 Land system disaggregation accuracy assessment

From the initial field validation an overall accuracy of 74% was recorded. Land components 1–7 (96.8% of the study area) achieved an overall accuracy of 84%, while components 8–10 achieved an overall accuracy of 50%. Table 5 summarises the results of the field validation exercise. The predominant classification inaccuracies involved an over‐classification of the ‘prominent lunettes’, ‘prominent former lakebeds’ and ‘lunettes associated with ridges’ landform classes. The average results (50%) for the ‘lunettes associated with ridges’ and the ‘prominent lunettes’ landform classes can in part be attributed to the large scale of these landform features and their relationship to other neighbouring landforms (e.g. ridges and depressions). These factors compounded the difficulty in differentiating these landforms from ridges and rises respectively in the validation assessment. These particular landform components require further soil and geological analysis to verify their differences and landform classification. Whilst it was recognised that many of the allocated ‘prominent former lakebed’ sample points were possibly lakebeds in the past, the nature of their topography did not clearly distinguish them from plains elsewhere in the landscape. It was decided that only the very obvious lakebeds would be confirmed whilst the others would be reclassified as undulating plains. In response to the validation results the following changes to the landform map have been made: • Mapped former lakebeds that were juxtaposed to mapped Raak Land System units and a few others that were clearly identifiable through a visual analysis of the DEM and air photographs were kept as ‘prominent former lakebeds’ and the remainder were changed to ‘undulating plains’. • Only lunettes associated with the refined lakebed landforms were kept whilst the others were reclassed to ‘rises on undulating plains’. • The ‘lunettes associated with ridges’ that were correctly validated were kept whilst others incorrectly classified were reassigned to ‘prominent ridges’. • The relative elevation threshold to disaggregate the ‘rises on ridges’ landform class was raised from 0.54 to 0.7. This was to ensure ridge crests and the tops of rises such as hummocks were mapped whilst other areas that may simply be part of the ridge slope were reclassed to ‘prominent ridges’. It is anticipated that the overall accuracy post the validation survey results and following modifications to the landform map would be significantly higher than 74% and possibly closer to 84% achieved for the land components 1–7.

27 DEPARTMENT OF PRIMARY INDUSTRIES

Table 5 Validation results for each landform class. Incorrect samples were assigned by visual assessment to the next most likely class

Sample results

Percentage of sample assigned to a different landform class

Landform

correct R PL RR UP PFL LRS % RUP LAR EWDP EWDR

East–west dunes on plains (EWDP) 89.7 3.4 6.9

East–west dunes on ridges (EWDR) 80.0 7.0 10.0 3.0

Prominent ridges (R) 80.0 20.0

Rises on ridges (RR) 76.7 3.3 20.0

Lower ridge slopes (LRS) 83.3 16.7

Rises on undulating plains (RUP) 86.6 3.3 3.3 3.3

Undulating plains (UP) 92.9 7.1

Prominent lunettes (PL) 56.7 6.7 30.0 3.3 3.3

Lunettes associated with ridge (LAR) 46.7 53.3

Prominent former lakebeds (PFL) 46.7 53.3

Total (Average) 73.9 1.7 0.3 3.0 1.0 0.0 3.6 15.7 0.0 0.3 0.0

3.4 Landform pattern regions

Whilst there have been a number of efforts to delineate homogenous areas at a finer scale than the original land system mapping, this is the first opportunity to do so based on mapped landform. This delineation recognises that within the existing extents of the land systems there is variability in landform patterns. The broader landform pattern mapping also assists in distinguishing the morphological variation in land formations that have been classified as the same landform component. For example, what has been mapped as rises on plains in one region may have different relief characteristics than the same class in a different region of the land system. This delineation has been based largely on a visual interpretation of the map outputs created in this project and as such should be considered as a draft attempt that can act as a guide to further refinement. Jim Rowan’s refined land system mapping from 1987 (LCC 1987) and the latest geomorphological unit (GMU) mapping were also used to aid the creation of regional boundaries—although it is acknowledged that these classifications incorporate other factors such as climate and soil. Eighteen regions have been identified and mapped, 10 in the Central Mallee Land System and eight in the Hopetoun Land System. The isolated Hopetoun areas in the north have been classified into one of the Central Mallee classes as their position in the landscape and their landform pattern was not dissimilar to that of the Central Mallee (J Rowan pers. comm., 3/10/08). Table 6 describes the landform patterns contained within each region.

28 DISAGGREGATION OF MALLEE LAND SYSTEMS

Table 6 Landform pattern regions in the Central Mallee (CM) and Hopetoun (H) land systems

Code Description CM1 Prominent plains with moderate linear dune density and interdunal swales. Remnants of the ridges are present. CM2 Elevated plains with moderate linear dune density. Remnants of the ridges are present. CM3 Elevated plains with high linear dune density. Remnants of the ridges are present. CM4 Depression with low linear dune density. CM5 Prominent plains with high linear dune density and interdunal swales. Remnants of the ridges are present. CM6 Low‐lying plains with rises and low linear dune density. Rises may include remnant ridges and lunettes associated with former lakes. The rises here trend north–south and have a higher relief than in other areas of the Central Mallee. CM7 Prominent plains with high density converging linear dunes and interdunal swales. Remnants of the ridges are present. CM8 Prominent ridges and ridge slopes with moderate linear dune density CM9 Prominent plains with low linear dune density. Ridges remnants are mostly absent. CM10 Elevated plains with low linear dune density. H1 Prominent ridges and inter‐ridge features including linear dunes, former lake beds, lunettes, ridge slopes and undulating plains. Linear dunes become more frequent in the northern parts of this region. H2 Prominent level plains and poorly formed linear dunes. Ridge remnants are mostly absent. H3 Prominent undulating plains with remnant ridges and poorly formed linear dunes. H4 Undulating plains with moderate linear dune density and depressions associated with Lake Tyrell. Remnants of the ridges are present. H5 Elevated plains with rises and low linear dune density. Remnants of the ridges are present. H6 Prominent plains with low linear dune density. Ridges remnants are mostly absent.

3.5 Area analysis and statistics of the landform components

Some area analysis on the landform component distribution across the land systems and within each land pattern region has been performed (Table 7). Statistics on precision and accuracy of the outputs has been provided through the field validation (Section 2.7). A map showing the density of the linear dune landforms across the study area has been created and is presented below (Figure 23). Each full grid cell represents about 160 km2. This map and the area cover data (Table 7) correlate well with the GMU250 dataset (Geomorphology mapping of Victoria at 1:250,000 scale) and with Rowan and Downes’ (1963) original landform distribution description within the two land systems, and provides a guide to further refinement of the landform pattern regions (Figure 24).

29 DEPARTMENT OF PRIMARY INDUSTRIES

Figure 23 Percentage cover of the east–west linear dunes

Figure 24 Map showing the landform pattern regions and their area as described in Table 5 A noticeable possibility is the lack of prominent ridges in the eastern, western and isolated northern parts of the Hopetoun Land System. These areas stand out as being markedly different from the central core of the land system that is dominated by ridge and inter‐ridge features. MacEwan et al. (2007) proposed a revision to the extent of the Tempy Land System by refining it to the areas dominated by ridges—the principal

30 DISAGGREGATION OF MALLEE LAND SYSTEMS landform component as described by Rowan and Downes (1963). Similarly, regions of the Hopetoun Land System that lack prominent ridges may be better allocated to the Central Mallee or other adjacent land systems. The similarity in landform pattern between the refined Tempy and the H1 region suggests a merging of these two regions would be logical. Table 7 Area cover (expressed in percentage) of each mapped landform component in each landform pattern region (mapped in Figure 24).

slopes cover) ridges plains ridge

plains

former plains ridges area lunettes

ridges

(% on

on on cover) cover) Region cover) cover) cover) cover) cover) cover) cover)

ridge

(% (% (% (% (% (% (% (% (% (hectares)

Region Rises Dunes Dunes Prominent Prominent Prominent Undulating Lunettes lakebeds Prominent Lower

CM1 158,955 0.05 0.02 0.00 17.22 67.28 13.69 0.01 1.58 0.00 0.14

CM2 417,540 0.03 0.02 0.00 16.51 67.67 13.98 0.02 0.80 0.13 0.83

CM3 15,894 0.00 0.00 0.00 7.60 70.30 21.88 0.00 0.19 0.00 0.00

CM4 4329 0.00 0.00 0.00 10.43 42.20 6.20 0.00 36.33 0.00 4.83

CM5 33,164 0.00 0.00 0.00 9.34 72.45 17.44 0.00 0.68 0.00 0.09

CM6 134,311 0.14 0.21 0.00 24.60 55.06 8.91 0.11 6.22 0.00 4.76

CM7 12,568 0.00 0.00 0.00 10.32 72.86 16.40 0.00 0.42 0.00 0.00

CM8 92,890 22.99 10.51 0.00 12.10 42.57 5.66 5.11 0.45 0.59 0.00

CM9 122,462 0.00 0.00 0.00 26.63 59.01 9.55 0.00 4.21 0.00 0.60

CM10 3897 6.62 33.37 0.00 14.42 44.12 0.39 1.08 0.00 0.00 0.00

H1 256,549 15.40 14.59 17.68 27.96 14.91 0.97 1.55 1.36 0.92 0.62

H2 10,345 0.04 0.03 4.59 76.24 18.34 0.00 0.00 0.00 0.00 0.76

H3 33,886 1.76 1.78 0.00 64.29 31.44 0.53 0.06 0.14 0.00 0.00

H4 23,142 0.76 0.44 0.03 64.18 20.43 8.18 0.23 4.44 0.00 1.31

H5 33,060 0.00 0.00 0.00 27.16 68.00 4.84 0.00 0.00 0.00 0.00

H6 3559 0.00 0.00 0.00 70.50 24.27 4.28 0.00 0.95 0.00 0.00

31 DEPARTMENT OF PRIMARY INDUSTRIES

4 Conclusion

The improved resolution in satellite and aerial imagery has facilitated the disaggregation of individual landforms within each land system. The disaggregation process together with new data has led to the identification of landform patterns within the land system extent, such as the former lakebeds and associated lunettes and the high relief rises in the CM6 region in the western part of the Central Mallee Land System. The landform pattern mapping can inform the adjustment of land system boundaries to better reflect the types and distribution of landforms that exist within the Mallee region. Importantly, the Tempy/Hopetoun example shows that landform patterns will occur across existing land system boundaries and as such any alterations in the extent of land systems based on landform mapping can only be interim until all land systems within the catchment have been disaggregated into their landform components. It is also acknowledged that land system definition involves more factors than just landform (such as soil, vegetation and climate information) and therefore these factors will also need to be considered when making any amendments. As the landform components within a land system provide a basis for defining likely soil occurrence and associated land degradation issues, the landform mapping can inform the development of a targeted approach to soil and land monitoring, research trials and significantly supports whole farm planning programs such as eMAP. Finally, the landform dataset will provide a valuable input into the Land Use Impact Model (LUIM) to assess the likelihood of wind erosion in these two land systems. The increased resolution will complement the recent utilisation of remote sensed satellite data for evaluating ground cover resulting in a within‐ paddock scale assessment. This has already been successfully trialled in land use impact modelling of the Tempy Land System using the disaggregated landform dataset to evaluate land susceptibility to wind erosion (MacEwan 2007). The increased detail of the model outputs will allow for a more precise and targeted approach to land management priority settings.

4.1 Recommendations

The development of a high resolution landform dataset covering almost 40% of the Mallee region provides a number of opportunities for further investigation and research. 1. Process to complete land system disaggregation for the Mallee region including internal consistency checking and revision to align derived landforms with existing landforms defined for the Hopetoun, Central Mallee and Tempy land systems. Refinement of land systems mapping using radiometric data, field survey and recently acquired orthophotographic DEM (including integration of the VGF) would benefit the development of a nested hierarchy of soils and land mapping for the region. Refinement of land system boundaries and divisions (including disaggregation of remaining land systems) would accompany this revision to reflect the latest landform mapping (type and distribution). Priority land systems include the Millewa, Boigbeat, Culgoa and Raak, followed by Lindsay Island, Neds Corner and Big Desert. In the future this refinement (with development of a landform map base) could include assignment of critical soil properties to soil‐landforms with regard to monitoring priorities including wind erosion, soil structure decline, salinity and water erosion. In addition, updating this mapping in co‐operation with South Australian datasets will enable consistent soils and land data across states, enhancing capacity for cross‐ state research and development programs. 2. Revision of Rowan and Downes’ (1963) land report for north‐west Victoria. This was highlighted under the Victorian Mallee Soil Conservation Action Plan background report (Cummins 2005) as Rowan and Downes report ‘provides an important resource manual for extension providers, educators and land managers’. Coupled with a refinement of land systems for the Mallee region,

32 DISAGGREGATION OF MALLEE LAND SYSTEMS the revision would integrate new research and soils data into descriptions to fit the growing needs of potential users. Another consideration is the ongoing support and enthusiasm for this work that has been provided by Jim Rowan, which will continue into the future. 3. Collation of existing critical soil and landscape parameter properties including land use, management history and land degradation into the Victorian Soil Information System and Victorian Land Use Information System. Collation of soil site information existing in a variety of states (e.g. hardcopy report, digital datasets, historic sites including soil pits from various research and extension activities, agronomic trial sites) that are currently absent from the Victorian Soil Site Database. Harnessing this soil information data will enable a suite of stakeholders across the Victorian Mallee region to utilise best available soils data in decision marking and research and development planning. 4. Acquire critical soils data towards fulfilling national soil condition reporting requirements and management action targets. Here reference sites could be established for major soils of the Victorian Mallee region with ongoing measurement of critical soil properties to fulfil monitoring objectives.

33 DEPARTMENT OF PRIMARY INDUSTRIES

References

Cummins T (2005) Soil Conservation Action Plan. Mallee Catchment Management Authority. Gallant JC, Dowling TI (2003) A multi‐resolution index of valley bottom flatness for mapping depositional areas. Water Resources Research 39(12), 1347–1359. LCC (1987) ‘Report on the Mallee area – review.’ Land Conservation Council, Victoria. MacEwan R, Clark R, Robinson N, McNeill J (2007) ‘Assessing wind erosion risk in the Mallee.’ Department of Primary Industries, Victoria. Murphy B, Vaze J, Teng J, Tuteja NK, Gallant J, Summerell G, Young J, Wild J (2005) Modelling landscapes using terrain analysis to delineate landforms and predict soil depths – examples from catchments in NSW. In ‘Proceedings of the MODSIM 2005 International Congress on Modelling and Simulation’. (Eds. A Zerger, RM Argent) (Modelling and Simulation Society of Australia and New Zealand). 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. URL http://www.clw.csiro.au/publications/technical/technical97.html. Robinson NJ (2008) ‘A terrain analysis assessment of waterlogging susceptibility for the Corangamite CMA region.’ Department of Primary Industries, Victoria. Rowan JN, Downes RG (1963) ‘A study of the land in North‐Western Victoria.’ Soil Conservation Authority, Technical Communication No. 2, Victoria. Summerell GK, Dowling TI, Wild JA, Beale G (2004) FLAG UPNESS and its application for mapping seasonally wet to waterlogged soils. Australian Journal of Soil Research 42(2), 155–162. Summerell GK, Vaze J, Tuteja NK, Grayson RB, Beale G (2005) Delineating the major landforms using an objective hydrological terrain analysis method. Water resources Research 41, W12416, doi:10.1029/2005WR004013.

34 DISAGGREGATION OF MALLEE LAND SYSTEMS

Appendix 1 Field validation route map

35 Mallee Catchment Management Authority www.malleecma.vic.gov.au PO Box 5017 Mildura 3502 Telephone 03 5051 4377 Facsimile 03 5051 4379