1996/97 Land Use of Australia. Final report for project BRR5.

Prepared by J.B. Stewart, R.V. Smart, S.C. Barry and S.M. Veitch
December 2001

Citation:
Stewart, J.B., Smart, R.V., Barry, S.C. and Veitch, S.M. (2001)1996/97 Land Use of Australia - Final Report for Project BRR5 , National Land and Water Resources Audit, Canberra.

Acknowledgments

The authors would like to acknowledge the various contributors to this project. Paul Walker, Warren Muller, Shane Cridland, Kim Malafant and Dawn Fordham provided key inputs to the project in the form of data analysis, programming and advice and their documentation is reproduced and acknowledged in Appendices 1, 5 and 6. A number of staff from State and Territory agencies contributed control site data. In particular we would like to thank the coordinators in each State or Territory:

NSW: Ian McGowen (NSW Agriculture)

NT: Stephen Hester (NT Department of Lands Planning and Environment)

QLD: Darren Springer & Deanna Weller (QLD Department of Natural Resources)

SA: Richard Williams (Primary Industries and Resources SA)

TAS: Robin Thompson (Tasmanian Department of Primary Industries, Water and Environment)

VIC: Rhonda Day (FarmStats Australia Pty Ltd) & Andy McAllister (Agriculture, Victoria, DNRE)

Thanks also to the various project managers that supported this task and to Ag WA for provision of pasture control sites. The control site data was only available with input from landholders and so a big thank-you is due to the hundreds of landholders who participated.

A number of Bureau of Rural Sciences staff contributed GIS expertise and/or advice during the course of the project. We particularly acknowledge the inputs of Josh Cosgrave, Dale Young, Lisa Guppy and Jim Walcott.

Executive Summary

Introduction

The 1996/97 Land Use of Australia, Version 2 provides an area representation of dominant land use by level of intervention or intensity of land use. For agricultural land uses, the land use shown is that predicted to have occurred in 1996/97. The actual land use may differ from year-to-year reflecting crop and pasture rotations and use of irrigation.

Agricultural land uses were determined through an automated process to spatially allocate agricultural census data (AgStats 96/97, Australian Bureau of Statistics) using satellite imagery. Advanced very high resolution radiometer (AVHRR) data was processed to provide maximum NDVI (normalised difference vegetation index) composite images. A fortnightly sequence of NDVI images for the same period as the agricultural census was used to assign the agricultural land use. This was achieved by comparing the NDVI profile of a pixel with the NDVI profile of a known land use or control site. Those pixels with NDVI profiles most like the control site were assigned the control site's land use. The number of pixels assigned a particular land use was constrained to the area reported in the agricultural census for the region being solved. The method used is described in Walker & Mallawaarachchi (1998).

The control sites were provided by various state and territory agencies largely through field visits and farmer interviews. The participating agencies were: NSW Agriculture, Victorian Department of Natural Resources and Environment, Qld Department of Natural Resources, Primary Industries and Resources SA, Agriculture WA, Tasmanian Department of Primary Industries, Water and Environment and NT Department of Lands Planning and Environment.

Several available data sets were utilised to determine non-agricultural land uses. These were:

Collaborative Australian Protected Areas Database - CAPAD97 for Tasmania and CAPAD99 for mainland Australia (Environment Australia, 1998 and 2000 respectively) for protected areas.
GEODATA TOPO-250K, Series 1 (AUSLIG, 1999) for topographic features (water bodies, built-up areas and licensed airports).
Australian Tenure (BRS, 1997) for distribution of public, private and aboriginal land in Australia.
Native Forest and Plantations of Australia, 250m (BRS, 1999) for native and plantation forests and the 1995 Land Cover Theme: Queensland (BRS, 2000) to improve Queensland plantation data.

The data set is available for download from the Australian Natural Resources Data Library (http://adl.brs.gov.au).

Key findings

In conjunction with existing national data sets, the SPREAD (SPatial REallocation of Aggregated Data) method of Walker & Mallawaarachchi (1998) has enabled the compilation of the first national land use map in digital form. The features of the resulting 1996/97 Land Use of Australia data set are:

It is a nationally consistent product.
It is a spatial benchmark for land use in Australia circa 1996.
It has a flexible data structure to enable users to modify the land use classification to suit their requirements.
It has reliability indicators for the agricultural land uses mapped.
It embodies a spatially corrected version of the commodity area data from the 1996/97 AgStats database.

Content of the data set

From the 1996/97 Land Use of Australia data set the following can be stated regarding Australia's land use:

The most extensive land use is 'livestock grazing' covering 430 Mha or 56% of Australia.
Agricultural land represents 473 Mha or 62% of the continent.
About 2.2 Mha are irrigated for agriculture, representing 0.5% of all agricultural land or 5% of the area under crops, horticulture and modified pastures.
Cropping (dryland and irrigated) covers 22 Mha or 3% of Australia with cereals representing 18 Mha.
73% of the 323 000 ha under horticulture (perennial and seasonal) are irrigated.
Nearly 47 Mha or 6% of Australia has the land use, 'nature conservation', with 85% of this area having the gazetted IUCN categories strict nature reserve and national park.
'Water' occupies 13 Mha or nearly 2% of Australia, with 60% of this being water features with a conservation classification.
13% or 100 Mha of Australia is 'managed resource protection' with traditional indigenous uses being 89% of this area.
'Other minimal use' occupies 118 Mha or 15% of the continent. Reserved, vacant or institutional crown land represents 77% of this area and remnant native cover on private land occupies nearly 28 Mha.
Forestry (production and plantation) covers 2% of the continent.
'Intensive uses' occupy about 2.4 Mha, or 0.3% of Australia. 'Intensive uses' includes urban and periurban areas and open-cut mines.

Limitations of the data set and future improvements

Limitations of the 1996/97 Land Use of Australia data set are listed below. These limitations offer opportunities for improvement in future implementations of the methodology used.

A number of minor intensive land uses, most notably rural residential land and mining, proved difficult to distinguish from agricultural land.
Fine-scale spatial heterogeneity of agricultural land use is not handled well by SPREAD (i.e. where the typical land use parcel size is significantly less than the pixel size).
Fine-scale temporal heterogeneity of agricultural land use is not handled well by SPREAD (i.e. where the typical cropping period is six months rather than one year).
The matching of AgStats area data to mapped areas of agricultural land requires assumptions and adjustments based on the assumptions.
The ability of SPREAD to discriminate between the NDVI profiles of target zones limits the accuracy of the agricultural land use allocations.
The performance of the heuristic implementation of SPREAD is poorly understood.

Applications

The 1996/97 Land Use of Australia has many applications relevant to national scale issues. At a spatial resolution of 1:1 000 000 to 1:2 500 000 the national land use map is one of the National Land and Water Resources' fundamental data sets. It has been widely used by other Audit projects as a key input or for context setting. The data set will be useful in:

Providing primary attribute maps as background information products to inform decision-making and assist in the presentation of policy proposals (e.g. pair-wise comparisons Ð i.e. land use by soil type; combining appropriate tenure classes, protected areas, forests etc).
Supporting information on diversity of environment and related patterns in economic use of land (e.g. risk assessment of land use with climate reliability projections; relating land use to social factors).
Strategic industry-based planning to take advantage of increasingly deregulated domestic and global trading or to decide where to locate a new factory of office.
Integrating with other data sets to allow multi-objective assessment on land use and land use change (e.g. ecological services design, trade-off and sensitivity analysis, land use competition futures).

Each product should be supplemented with explanatory notes. The suggested approaches require statistically and spatially relating the land use data to other data sets and/or the integration of Audit themes.

Where more detailed land use mapping is available this will be useful in decision-making at the regional to catchment scale.

Some primary attribute products derived from the 1996/97 Land Use of Australia data set are provided on the Australian Natural Resources Atlas at www.environment.gov.au/atlas.

Recommendations

Implications for managing Australia's natural resources

In the management of Australia's natural resources:

The national land use map should be used as a spatial benchmark for land use in Australia circa 1996.
Users should make use of the data set's flexibilty to modify the land use classification to suit their purposes.
Users should consider the agricultural land use reliability indicators provided.
The data set should be recognised as a source of spatially corrected agricultural commodity area data derived from the 1996/97 AgStats and used as such when the opportunity arises.
Use of the 1996/97 Land Use of Australia data set should be encouraged for appropriate purposes in government, industry, research and education.
Possibilities for integration of the national land use map with other thematic data relevant to natural resource management should be explored.

Improvements to the national land use map

To ensure the success of future implementations of the methodology and to improve shortcomings of the current implementation, there is a need for:

Better mapping approaches for non-agricultural rural land uses, particularly rural residential land and mining.
A solution to mapping areas where land use parcel sizes are significantly smaller than the pixel size. A possible solution may be to use finer resolution imagery with smaller pixels or, at least, to do this in selected regions.
A solution to mapping areas where the characteristic cropping period is only six months rather than one year. One possible solution might be, in the implementation of SPREAD, to characterise phenology over two consecutive six monthly periods, independently.
Regular collection of agricultural census data. The potential release of geocoded agricultural census data should be encouraged. Geocoding offers a way to identify and quantify errors and facilitate spatial correction of the data and to facilitate the identification of ground control sites for the SPREAD method.
A review of the agricultural census questions with the view to including supplementary questions to improve SPREAD's assignment of agricultural land uses.
Investigate the possibility of Improving SPREAD's capacity to discriminate between the NDVI profiles of target zones. Options include increasing the number of control sites, using a better discriminating metric and grouping different commodities into more easily discriminated groups.
A better understanding of the heuristic implementation of SPREAD and its limitations.

Choice of land use methods and suggestions for the future

Some considerations for future land use mapping are as follows:

The investment of the NLWRA in the Land Use Mapping of the Continent Using AVHRR Data project, resulted in a land use map at a scale ranging from 1:1 million to 1:2.5 million, costing approximately 12.5 cents km-2.
To re-run the AVHRR-based method is estimated to cost between 3.3 and 13 cents km-2.
Detailed land use mapping done for the NLWRA, resulting in land use maps at scales ranging from 1:25 000 to 1:250 000, cost approximately 50 cents km-2. The finer detail and potentially greater accuracy of ground-based and related methods has the benefit of being more reliable for decision making in small regions and catchments.
New technologies will be beneficial, such as the global positioning system, to enable convenient new ways to conduct surveys.
Geocoded agricultural census data would provide a source of ground control sites for the AVHRR method and enable it to be re-run with high levels of confidence, particularly in areas where land use parcel sizes are large compared with the 1 km AVHRR pixels. High resolution imagery, such as Landsat TM, would enable the same methodology to be used with improved results in areas where land use parcel sizes are small.

Acronyms

Organisations

NRIC	National Resource Information Centre
BRS	Bureau of Rural Sciences
CSIRO	Commonwealth Scientific and Industrial Research Organisation
ABS	Australian Bureau of Statistics
NHT	Natural Heritage Trust
ERIN	Environmental Resources Information Centre
EA	Environment Australia
AUSLIG	AUstralian Surveying and Land Information Group
NFI	National Forest Inventory
ABARE	Australian Bureau of Agricultural and Resource Economics
IUCN	International Union for Conservation of Nature and Natural Resources
PSMA	Public Sector Mapping Agency

Terminology

GIS	Geographical Information System
SRIAS	State Resource Information and Assessment System
AVHRR	Advanced Very High Resolution Radiometer
NDVI	Normalised Difference Vegetation Index
SPREAD	SPatial REallocation of Aggregated Data
SLA	Statistical Local Area
ASGC	Australian Standard Geographical Classification
CAPAD	Colloborative Australian Protected Areas Database
ALUMC	Australian Land Use and Management Classification
AGD66	Australian Geodetic Datum 1966
WGS84	World Geodetic System 1984

Australian Natural Resources Atlas

Natural Resource Topics

1996/97 Land Use of Australia. Final report for project BRR5.

Contents

Acknowledgments

Executive Summary

Acronyms