Soils - ASRIS

What is the Australian Soil Resources Information System (ASRIS)?

Introduction

ASRIS is a national database of soil information, suitable for use at national to large regional scale. It provides for Australia's agricultural zone (see explanation and map below):

1. Digital maps of key soil properties (estimated using models) on a 1.1 square km grid.

2. A national database of existing primary data relating to soil and land resources, comprising :

soil profile database
soil and land resources map database
other datasets relevant to modelling soil properties

What is the ASRIS study area?

ASRIS databases cover catchments (river basins) designated by the State agencies as containing significant agricultural activity. Catchment boundaries were those defined by AWRC (http://www.auslig.gov.au/meta/meta5.htm). The entire catchment was included (rather than just the area of agricultural activity) because the modelling process involved use of terrain descriptors that relate to position within the catchment.

Some soil properties were mapped over the entire Australian continent - these were limited to those soil characteristics interpreted from soil maps.

ASRIS study area

Who was involved?

Various agencies and organisations participated in the development of ASRIS including:

How was ASRIS built?

Land resource assessment and soil survey underpin sustainable land management. The Audit initiated the development of the Australian Soil Resources Information System (ASRIS) to compile information on the key soil properties that contribute to productivity, soil resilience and key processes controlling water, air and nutrients.

ASRIS is a national database of information on Australia's soils. It was designed to provide nationally conformable information from the extensive soil point and soil survey map data that has been collected by the State and Territory agencies and CSIRO. It contains both primary data (soil point and soil survey map data) and modelled estimates of soil properties based on these data.

Until the development of ASRIS, the best available digital coverage of soil information for all of Australia was the Atlas of Australian Soils. This was compiled in the late 1960's at a scale of 1:2,000,000 which provides only the broadest national overview of soil attributes. Since then, more detailed soil data have been collected by Commonwealth, State and Territory agencies. . Much of this information has been collected over the last 10 years through national programs such as the National Landcare Program and more recently the Natural Heritage Trust.

ASRIS was developed as a joint project between the Bureau of Rural Sciences, CSIRO Land and Water and the State and Territory agencies responsible for soil and land management. ASRIS built on the collaborative framework established by the Australian Collaborative Land Evaluation Program (ACLEP). The Working Group on Land Resource Assessment with representatives from the Commonwealth and each State and Territory formed the Steering Committee for ASRIS.

ASRIS therefore:

uses the best available information from State and Territory databases.
makes spatial estimates of soil properties (and their uncertainties) for the more intensively used agricultural zone of Australia.

The soil attributes estimated are those most commonly required to characterise, model or predict land resource processes that drive plant productivity, measure resource sustainability or control rate of resource degradation.

ASRIS contains:

a compilation (from data held by Commonwealth, State and Territory agencies) of soil profile data into a single database containing over 160,000 profiles in a standard format (SITES). These data are available from the Audit data library, subject to some licence conditions.
a compilation (from data held by Commonwealth, State and Territory agencies) of soil and land resources maps at varying scales. These data were used in modelling, and descriptions of the data are available from the Audit Atlas and Data Library. The data can be obtained only from the original custodians.
various ancillary datasets relevant to soils and used in modelling soil properties, including: 9 second DEM and derived terrain attributes; lithology (derived from geological mapping); climate surfaces; Landsat MSS.
a set of spatially distributed estimates of soil attributes and their data quality, in the form of gridded (raster) maps of soil properties for topsoil and subsoil. These maps were produced from the collated datasets using several different modelling methods.

The soil resource information can be used for the measurement, modelling, and prediction of land degradation impacts, system sustainability, and productivity.

Modelling of soil properties

Establishing sustainable agricultural and land management systems requires information not just on the distribution of soil types, but on the spatial variation in soil properties - for example, soil texture, water holding capacity and soil fertility. An important objective of ASRIS was to produce nationally consistent spatial estimates of key soil properties, suitable for use in regional to national scale assessments.

The collated ASRIS datasets were used as inputs to models to estimate soil properties, based on:

1. point-based models, used when sufficient soil profile data was available to build reliable models; or
2. polygon-based models, used when soil profile data were limited; or
3. combined point and polygon-based models.

Modelled estimates were produced for

layer 1 (topsoil) - the first or thickest A horizon; and
layer 2 (soil) - usually the first B or subsoil horizon.

Estimated soil properties are presented as grids (surfaces) with a cell size of 0.01 degrees in longitude and latitude (approximately 1.1 kilometres). Most of the input datasets were at 1:250,000 scale or finer, with the exception of the climate surfaces which were based on 5 km² grids.

It is important to stress that the national / large regional scale information presented in ASRIS is not suitable for use at the scale of small catchments or individual farms, as it is generalised to provide the best representation over large areas.

The table below sets out the soil properties modelled and the method of estimation for each. Click on the required property to display a digital map (at national scale), a description of the property and its significance, and details of how the map was produced.

Point-based models

New methods were developed by CSIRO to predict each soil attribute individually, on the basis of correlations between soil properties from soil profile data, and other environmental variables. Point-based models are based on the assumption that soil properties vary as a result of environmental variables such as climate and topography as well as soil type.

(Figure) is a schematic representation of the method used for the point models. For each soil property, a matrix was compiled of environmental variables against the soil property being predicted for each point, from which regression and classification models can be built. Since the environmental variables are available over the whole area, it is possible to use these models to extrapolate to areas where no measurements have been made, on the basis of their environmental similarity to points with measurements.

Binary decision trees were built for categorical data (such as texture), using the software modelling package C5.0. Piecewise regression models were built for continuous data (such as pH), using the modelling package Cubist.

The environmental variables used in the modelling included:

climate (19 surfaces)
terrain attributes (eg elevation, slope, distance to ridges/rivers etc, derived from the DEM)
lithology
Landsat MSS imagery
land use
soil type by Australian Soil Classification, from the Atlas of Australian Soils.

These predictor variables operate on different scales - for example MSS was initially available at a 100 metre resolution, the climate surfaces at a 5 kilometre resolution and the digitized ASC on a 1:2.5 million scale. This is not seen as a problem, and may in fact be advantageous given soil?landscape processes are known to operate over a range of scales, and consequently variation in soil properties may be explained by different factors at short and long range scales.

Reliability of point models

Modelled surfaces are presented at a grid resolution of 1.1 km² (0.01^o). Most of the input datasets were at 1:250,000 scale or finer, with the exception of the climate surfaces which were based on 5 km² grids.

All models were validated by separating the data into a training set and a test set. Model performance was also assessed by cross-validation on the full data set. Common error diagnostics used to assess the reliability of the model are described below.

Each soil property map is accompanied by a matching grid, representing the reliability of the modelled values.

Two major sources of uncertainty associated with point-based models are:

the predictive power of the models (that is, the degree to which relationships could be established between each soil property and environmental predictors)
the degree of extrapolation, which is governed by the local point density and how well that environment is represented in the database.

Generally, the more data available, the more robust the model and the less uncertainty in the predictions, because there is less extrapolation into areas / environments which are not well represented in the model construction stage.

For most soil properties, there are areas with very few soil points with measured data to support predictions. This is particularly true of the following regions: Northern Territory, Carpentaria, Far North Queensland, western South Australia and northern Western Australia.

Error diagnostics

Some commonly used error diagnostics presented with each digital map are:

N	the number of points used in the model
R²	an estimate of the percent of the overall variation in the property explained by the model
RMSE	(Root Mean Square Error): This is an estimate of the standard deviation of the errors. A lower RMSE is associated with greater predictive ability. RMSE values can not however be compared between for different properties because they depend critically on the scale used.
Correlation	the strength of the linear association between the observed and the predicted values.
Absolute error	This is the average absolute difference between the observed and predicted values. Lower absolute errors tell us that the predicted values are closer to the observed values more often. Absolute errors do however depend critically on the scale of the units used and so can not generally be compared between models.
Relative error	This is defined as the ratio of the average absolute error magnitude to the the error magnitude that would result from predicting the mean value. If there is little improvement on the mean the environmental variables have little predictive capacity and the relative error is close to 1. Generally the smaller the relative error the more useful the model.

Polygon-based models

Map (polygon) based models are based on the assumption that soil properties can be predicted directly from soil type. Polygon based models were used to model those soil properties (particularly soil water attributes) for which insufficient soil profile data were available to produce reliable models based on the point data.

Maps of soil type are linked to look-up tables of soil properties, following the method described by McKenzie et al (2000). Dominant soil types in each map unit were identified, the interpreted values for the relevant soil type were weighted by an estimate of the area occupied by each soil type, and the weighted value was ascribed to the map unit.

Where only the dominant soil type for a map unit was recorded, the estimate of each soil property for the entire unit was the value for that single soil type. Where a number of soil types and their relative proportions were recorded for each map unit, the soil property for the map unit was estimated as a weighted mean of the values for the soil types occurring in the unit.

detailed description of the polygon-based modelling methods can be found with each digital map.

A schematic representation of the polygon-based modelling process is shown below.

Click here for the larger image

ASRIS polygon datasets used in map-based models are set out in the table below. The scale of the various soil maps used in the polygon modelling is shown in figure below.

Datasets used in polygon-based models

Area	Map data	Soil property data
Western Australia	WA soil landscape mapping (system and subsystem, scale 1:100,000 - 1:250,000)	Tables of soil properties for WA Soil Groups, provided by Agriculture WA
South Australia	SA soil landscape mapping (scale 1:25,000 ? 1:100,000)	Tables of soil properties for SA Soil Groups, provided by PIRSA
Other states and territories	Best available mapping with Northcote Principal Profile Form assigned to map units (scale 1:50,000 ? 1:2 million).Where no other data were available, the Atlas of Australian Soils was used.	Interpretative tables of soil properties for Northcote PPFs, compiled by McKenzie et al (2000)

Reliability of polygon models

Uncertainties associated with polygon-based models are related to:

scale and accuracy of the base maps. The map above indicates the scale of the polygon data used in the modelling
how well the maps represent soil variability - for example, whether one dominant soil or a range of soil types is recorded for each map unit.
variation in the accuracy of tabulated estimates of the different soil properties
whether estimates of soil properties existed for specific soil types. McKenzie et al (2000) were not able to estimate all soil properties for all principal profile forms. For ASRIS, some additions were made to the McKenzie et al (2000) look-up tables, using data from the ASRIS point database)

A brief discussion of reliability is included with the description of each digital map.

Combined point- and polygon-based models

Point- and polygon-based models were combined in two different ways:

1. A digital map derived from a polygon model was included as one of the environmental layers in a point model. For example, the estimates of %clay derived from a polygon-based model were included as one data layer in the point model of %clay. This has the advantage of retaining the spatial structure of the soil map, while allowing estimates of %clay to vary within each map unit as a function of other environmental predictors. In this case, the errors in the final map are dominated by the errors in the point-based modelling process, since the polygon model surface is only one input to the point model.
2. Digital maps derived from different methods can be combined to give estimates of another property. For example, the digital map of soil erodibility was estimated using a pedotransfer function based on ? a measure of soil texture derived from polygon-based models of %clay, %silt and %sand
organic carbon derived from a point-based model permeability classes derived from polygon-based model of saturated hydraulic conductivity.

Here, the errors associated with the polygon model will dominate, since theses are generally larger.

What has been achieved by developing the Australian Soil Resources Information System?

Prepared 27 new data sets of key soil attributes using the best available information
Demonstrated the use of soil point and polygon data for spatial modelling and prediction
Facilitated quality assurance procedures on Australia soil information bank thereby improving the information base.
Implemented and refined the Soil Information Transfer and Evaluation System (SITES) protocols
Tested Australia soil point data for compliance with published standards for soil description and laboratory test data
Identified data quality issues such as geo-referencing, sample bias and lack of representativeness in the existing data bank
Identified existing significant data sets that are not part of the national data bank
Highlighted the impact of changing standards on data utility, for example, a change in the national standard for soil classification.
Initiated the development of a polygon database and transfer standard ? Developed a range of methods, appropriate to the availability and accuracy of the input data, to estimate soil attributes
Developed sophisticated set of spatial analysis procedures and tools for handling a mixture of point and polygon (taxonomic) data. These methods have application beyond soil science.
Developed tools for assessing the quality of the spatial predictions ? A tool set for preparing soil attribute maps
A national database of quality soil point data for future applications
Institution arrangements, through a partnership of Commonwealth, State/Territory agencies, to maintain, update and develop new applications with ASRIS

Maintaining our investment - Next steps

Building the Australian Soil Resources Information System was only possible because, in most cases, assessors of land resources have been collecting land (including vegetation) and soil data in the field to standard attributes and definitions. This has resulted in the collection of comparable data and information across Australia.

Data about soils continue to be collected. If we wish to improve our capacity to answer questions about the soils supporting agricultural landscapes, it is essential that new data are also consistent and comparable Australia-wide.

Recognising the need for maintenance of standards, the Australian Collaborative Land Evaluation Program (ACLEP) was established in 1992 to coordinate the development, update and review of technical standards for land resource assessment. The continued development of data standards, tools for analysis and reporting are particularly important for technical data if these data are to be used by non-specialists.

By the mid-1990's ACLEP in conjunction with Western Australia and Queensland developed the Soil Information Transfer and Evaluation System (SITES) in response to a growing demand for land resource information across agencies and Australia. SITES is a data exchange protocol for soil point data. SITES provided a standard for the exchange of data between all State/Commonwealth agencies - without these standards, ASRIS would not have been feasible within the timeframe.

Technical standards need to be extended and maintained as the demand for comparable, quality assured data continues to grow particularly at the regional scale.

95% of the ASRIS project was spent on "cleaning up" data. Inadequate data in mandatory or important data fields, and inconsistent adherence to data standards remain a significant issue for Australian land resource databases.

An essential activity for the national coordinators of ASRIS is to ensure the implementation of technical standards so that the quality of the input data to ASRIS over time is enhanced.

ASRIS analysis and reporting tools will need to be (re)-developed and maintained to improve the use of the ASRIS data outside of the technical practitioners within research and government agencies.

These tools could take the form of user guidelines or, for more advanced applications, computer aided decision support systems that use the underlying ASRIS.

Before committing to ongoing development of the analysis tools it is important that there is a comprehensive understanding of user needs, and that there is commitment to maintain any tools developed.

Continuous improvement in the development and adoption of technical standards in natural resource assessment should be encouraged. However, it is important that the considerable legacy of soil data is not alienated by the adoption of technical standards that do not maximise the use of this historical data.

A case study of data alienation

Soil classifications are designed to communicate an understanding of the attributes important to soil use and/or behaviour. The Atlas of Australian Soils (1968) used a classification called the "Factual Key".

In 1996, the Australian Soil Classification was adopted as the Australian standard. This classification does not yet have interpretive tables that allow data collected using the Factual Key to be matched to the Australian Soil Classification. From mid to late 1990s several resource assessment agencies stopped using the Factual Key in their mapping program.

The implication for the development of ASRIS was that the data collected using the different classifications could not be combined for the assessments undertaken by the Audit.

How can ASRIS be used?

Outputs from ASRIS have already been road tested. ASRIS attributes have been used to prepare comprehensive assessments of water-borne erosion and river nutrient transport, the current and projected extent of soil acidification and landscape productivity.

Sediment and nutrient delivery

In their natural state, Australian soils are generally shallow, infertile and have a fragile structure, which makes them prone to the erosive forces of wind and water, when the soil surface is inadequately protected by vegetation cover.

A major issue in Australian land management is waterborne soil erosion and the consequent impact on land and water resources. Off-site, the effect of soil erosion reduces water quality in streams and water storages due to increases in sediments and associated nutrients and the consequent impacts on ecosystem health.

Knowledge about the sources and rates of erosion under past and present conditions is essential for understanding the consequence of changes in land use and climate on soil erosion.

The Audit assessment of waterborne soil erosion used ASRIS soil erodibility (using several primary input attributes) and total phosphorus data to model sediment and nutrient delivery from hillslopes to streams and ultimately estuaries. See the Waterborne erosion and Riverine nutrient transport information products for more detail.

Acidification

Soil acidification risk is a major agronomic sustainability issue because it insidiously reduces productivity. Clearing land for agricultural purposes accelerates soil acidification. In agricultural ecosystems, actions that break the natural carbon and nitrogen cycles cause acidification. The processes include:

removal of agricultural products which remove alkalinity
accumulation or depletion of organic matter containing alkalinity
leaching of nitrogen which leaves acidity in the soil
addition of ammonium based fertilisers

The soil acidification process is primarily controlled by soil type, land use and management. The Audit developed a spatially explicit acidification risk model based on ASRIS attributes including the distribution of acid soils (pH), capacity for soils to buffer against acidifying practices (organic carbon % and percent clay) and plant yield functions (soil pH and other accessory characteristics). See the Soil acidification information product for more detail.

Landscape productivity

To manage Australia natural resource base in a sustainable manner requires knowledge and information about the spatial distribution of stores of nature's raw materials - carbon, water and nutrients - the major processes that control these balances and how these balances change in response to land use.

To model the cycles of carbon, nutrients and water in the landscape information is required that quantitatively and spatially defines terrain, climate, land use and soil stores for water and nutrients. Soil organic carbon and available phosphorus attributes were incorporated into the BIOS model that generated estimate of the soil and litter pools of nutrients and water. See Landscape balances information product for more detail.

Further Information - accessing ASRIS datasets

Australian Collaborative Land Evaluation Project (ACLEP)

The Australian Collaborative Land Evaluation Project encourages sustainable land use and environmental protection by promoting better procedures for acquiring and using land resource information amongst community groups, private industry and government.

ACLEP aims to set national standards for all aspects of land resource assessment, and encourage the development and application of innovative methods.

ACLEP reports to the Working Group on Land Resource Assessment (WGLRA), a steering committee of State and Territory land resource assessment agency representatives. The WGLRA also sets priorities for ACLEP and provides a formal link to the Commonwealth - State committee system via the Sustainable Land and Water Resource Management Committee (SLWRMC).

ASRIS has adopted many ACLEP initiatives and standards. The Working Group on Land Resource Assessment also formed the Steering Committee for ASRIS. Find out more about ACLEP here.

Working Group on Land Resource Assessment

With representatives from Commonwealth, State and Territory land resource assessment agencies, the Working Group on Land Resource Assessment (WGLRA) acted as Steering Committee for the project. The Working Group is a sub-committee of the Sustainable Land and Water Resource Management Committee (SLWRMC), reporting to the Standing Committee on Agriculture and Resource Management (SCARM).

For more information on WGLRA, including membership, click here.

Links and technical documentation

ASRIS modelled surfaces of soil properties are available for direct download from the Australian Natural Resources Data Library.

These are also available for on-line mapping.

Soil profile data compiled in ASRIS remain the property of the original custodian (State and Territory agencies and CSIRO). Soil profile data (in SITES format) are available for direct download from the Australian Natural Resources Data Library, with the written agreement of the data relevant custodian.

View a description of the point modelling method

Soil map data compiled in ASRIS remain the property of the original custodian (State and Territory agencies). Because these maps were compiled for a specific purpose, and because soil and land resources mapping programs are on-going, the compiled maps do not necessarily represent the best available information for a given area. For this reason, it is advisable to consult the relevant agency. Soil map data are available only from the custodian.

View a description of the polygon modelling method

Ancillary datasets compiled in ASRIS for modelling generally remain the property of the original custodians. For details of access conditions, follow the links to the sections on the relevant dataset

topographic descriptors and other digital elevation model (DEM) derivatives
lithology
climate
Landsat MSS imagery
land use

View a description of the ancillary data sets used to build the Australian Soil Resources Information System soil property maps

technical document detailing the development of soil property prediction using soil maps (PDF 9.1 MB)
technical document detailing estimation of particle size distribution (PDF 1.3 MB)
technical document detailing laboratory methods and correlation procedure (MS Word 260 KB)
spreadsheet detailing laboratory methods used and number of samples (MS Excel 86 KB)
table detailing land use classes used in Australian Soil Resources Information System modelling (MS Word 32 KB)
statement on data quality for South Australian data sets (MS Word 24 KB)
statement on data quality for Western Australia data sets (MS Word 24 KB)

Proposed Polygon Standards

technical document detailing the polygon data standard for land resource data sets (MS Word 77 KB)

technical document detailing the polygon attributes for land resource data sets (MS Word 24 KB)

Before you download

Most publications are downloadable as PDF files. Adobe Acrobat Reader is required to view PDF files.

If you are unable to access a publication, please contact us to organise a suitable alternative format.

Key

Links to an another web site
Opens a pop-up window

Key

Australian Natural Resources Atlas

Natural Resource Topics