Previous PageTable Of Contents

MaPPing agricultural commodities and land management practices from geocoded agricultural census data in Gippsland, Victoria

Lucy Randall and Michele Barson

Bureau of Rural Sciences,
PO Box E11, Kingston, ACT 2604
Phone: +61 (0) 2 6272-4901
Fax: + 61 (0) 2 6272-5825


Land management practices adopted by farmers have major impacts on natural resources outcomes. Changes in land management practices will make an important contribution to reducing Australia's salinity and water quality problems. Examination of the opportunities to modify land management practices and the likely outcome of such changes is seriously hampered by a lack of information on the spatial distribution of these practices.

This project investigated whether integrating catchment scale land use data with geocoded Agricultural Census data could be a cost effective way of producing accurate catchment scale land management and commodity information.

The Australian Bureau of Statistics (ABS) collected geocoded census data for Gippsland (Victoria) through the geocoding of farm cadastral boundaries. The geocoded data were rasterized and intersected with 1:100,000 land use maps prepared by the Victorian Department of Natural Resources and Environment (DNRE) to produce maps of commodities and land management practices. Data matching enabled mapping of 41 commodities for 73 percent of the geocoded area and tillage, fallow and stubble land management practices for 11 percent of the cropped area. Results indicate that the majority of broadacre crops were captured. Horticultural crops and pastures were less successful due to multiple cropping, particularly of seasonal vegetables, and native pastures under leasing arrangements were not captured in the geocoding exercise.

In order to ensure confidentiality of ABS' agricultural unit record data, no individual farm can be identified either by the farm boundary or commodity. This is maintained by the rasterizing and matching procedure and by grouping commodities into tertiary land use classes.

Methods for improving the quality of the resulting data have been identified including improving the coverage of the agricultural census, improving the census questions and response rate to land management questions and ensuring that land use maps are adequately field checked


The types of land management practices adopted by farmers have a major impact on natural resources outcomes (Lesslie, Barson and Randall, 2000). For example, water quality outcomes can be affected by choice of tillage methods, use of fallow, crop residue management methods, rates and timing of fertiliser applications and grazing management practices. Identification of the need for changes in land management practices to improve agricultural sustainability has led to establishment of the Murray-Darling Basin Commission’s Landmark Project. Such changes are also regarded as one of the major options for dealing with salinity issues under Australia’s recently established National Action Plan for Salinity and Water Quality Management.

Our ability to examine the opportunities to modify land management practices and the likely outcome of such changes is seriously hampered by a lack of information on the spatial distribution of these practices. A recent review (Lesslie et al, 2000) concluded that data currently collected on land management practices are of limited value. These data are only available in a highly aggregated non spatial format (Australian Bureau of Statistics’ Agricultural Census Statistical Local Area data), have been collected for a relatively small sample of farms (ABARE farm survey data) or are a one off, aspatial collection for a particular industry or region.

This project investigates whether geocoding the Agricultural Census could provide data suitable for mapping commodities and land management practices at catchment scale.

The Bureau of Rural Sceinces (BRS) and State agencies are currently producing catchment scale land use digital data sets (Barson and Lesslie, 2001) identifying 16 major classes of agricultural land uses as well as conservation and intensive land use classes (see for further details). However, the resources available for this mapping are insufficient to map the land management practices of farmers or the commodities they produce. Integrating catchment scale land use data with geocoded census data could be a cost-effective way of producing catchment scale land management and commodity information.

Study area

The Gippsland study area (Figure 1) comprises the Tambo, Mitchell, Thomson, Latrobe, East and South Gippsland, and Snowy catchments in Victoria. The region covers 3.86 million hectares and lies 70 kilometres east of Melbourne. The region supported a population of 223,300 in 1997, with employment being principally in agriculture and coal mining.

Figure 1. Location of the Gippsland study area

Methods and data sources

The overall procedure is shown in Figure 2.

Figure 2: Procedure for spatial matching of land use and agricultural census data

Geocoding of farm boundaries

ABS established from its Business Register that there were 5,912 farms in Gippsland in 1999. Maps showing land parcel boundaries were sent to farm owners or managers, who were asked to delineate the boundaries of the land parcels they were currently operating. About 88% responded, although 17 did not supply data suitable for geocoding (Australian Bureau of Statistics, 1999).

ABS digitised the farm boundaries into MapInfo and supplied the data as an Arc/Info shapefile. This shapefile was converted into an Arc/Info coverage. The coverage included address fields, farm area and most importantly the Integrated Register Identifier (IRID) number, which is the unique identifier for each unit record in the agricultural census. The total area geocoded was 815,352 hectares (21% of the study area) and comprised 4,843 farms or 82% of the farms reported in the 1996/97 census.

The location of the geocoded farms is shown in Figure 3. The majority of the farms are in west Gippsland, with some in the Omeo district, and the Cann and Snowy River valleys of East Gippsland.

Analysis of the agricultural census data

ABS has collected agricultural census data annually from the early 1990s until 1996/97. Collection of a core set of commodity data is funded by the ABS to estimate the value and area of agricultural production. These data items include the areas of holding, pastures, hay and silage, crops, orchards, and other fruit, grapevines, nursery produce, vegetables, numbers and sales of livestock and production of meat, wool and eggs. The 1996/97 census for Victoria also included questions on land management practices (stubble management, tillage and fallow practices and use of irrigation) funded by the DNRE.

The 1996/97 census data (AgStats, 1996/97) for commodities and land management practices were supplied for the Gippsland statistical local areas as two files:

  • census codes description
  • 1996/97 agricultural census data

Figure 3. Location of geocoded farms in the Gippsland study area

The latter comprised 5,912 unit records and 254 attributes including value (calculated by ABS) and production statistics (eg tonnes of wheat), neither of which were used in the analysis for this study.

The census area data were aggregated into broad commodity classes:

  • Cereals including oats, wheat, and millet
  • Non-cereal crops such as canola and field, mung and soy beans
  • Fruit and vegetables, (orchards and vines, soft fruit and vegetables)
  • Pastures (native and sown)
  • Livestock types, mainly sheep and cattle.

The dominant commodities and land use activities for each farm according to the census data were established. Dominant commodities and pasture types were identified as those occupying greater than 50% of the agricultural area of a farm for the following agricultural activities:

  • Cropping: cereals and non-cereals crops
  • Horticulture: fruit and vegetables
  • Grazing: pastures and livestock.

For some farms undertaking a diverse range of activities, it was difficult to determine the dominant commodity; in these cases, the commodities were classified as "mix" and the sub-dominant commodities were identified.

Land use data

DNRE supplied the Gippsland land use data as an Arc/Info coverage (Figure 4). Land use data was compiled from existing GIS data, satellite imagery classifications and cadastral data. Land use was classified according to the Australian Land Use and Management Classification (Bureau of Rural Sciences, 2001) to the secondary level for agricultural uses and at tertiary level for conservation and intensive uses (DNRE 2000). Land use was not mapped for the Woolami and Howitt mapsheets, west of Warragul, which were included in the ABS statistics.

Figure 4. Land use for the Gippsland study area

As shown by the geocoded census data, most of the agricultural activity is taking place in the south-western part of the study area. Significant areas of coal mining were found east of Warragul. Large tracts of State forest and conservation areas are found in the northern and eastern parts of Gippsland.

The land use data were summarised into broad land use classes that could be compared with the data generated from the geocoded census (Table 1).

Comparison of these two data sets shows that:

  • The area of cropping mapped is similar to the cropping areas reported in the census.
  • The area of horticulture mapped is about half that reported in the census; this is likely to be due to multiple cropping of soft fruits and vegetables.
  • The area mapped as grazing is almost exactly double that reported for the census. This was possibly due to inadequate reporting of native pastures in the agricultural census.

Table 1: Comparison of DNRE land use and agricultural census data 1996-97 for Gippsland

Land use

Area mapped (ha)

% of area mapped

Geocoded area (ha)

% of geocoded area
















Intensive uses




















n/a not applicable

Spatial matching of land use to agricultural census data

The DNRE land use data was matched to the geocoded farm boundaries within Arc/Info GIS. Matching took place both spatially as well as through attributes, by classifying the land use into broad agricultural uses, and then matching the dominant commodity on a farm-by-farm basis.

Analysis was carried out using Arc/Info 8.02. Coverages were converted into raster files or grids with a cell size of 25 x 25m based on the dominant land use in the grid cell. The land use data were merged using the COMBINE command in Arc/Info. Look-up tables linked the land code to the land use class, and the polygon identifier to the IRID and thus the census data. The analysis is summarised in Figure 5.

Figure 5. The procedure for matching land use to census data

Table 2 identifies the extent to which the areas of major land use classes mapped by DNRE were matched to the geocoded agricultural census data.

Table 2. Analysis of land use and geocoded agricultural census data matching

Agricultural land use

Secondary land use

Area of land use


Area of land use matched to geocoded data (hectares)

Match rate


Native pastures





Improved pastures
























Just under half the area mapped as agricultural land use was covered by geocoded census data. The areas of cropping were fairly well captured in the geocoded data. However, more than 80% of the area mapped as grazing of native pastures was not captured in the geocoding.

Investigation of the data showed that there was poor matching between areas of native grazing in the northern and eastern parts of Gippsland. These areas were compared with land tenure data. This comparison showed that 202,000 hectares mapped as grazing by DNRE were multiple use forests, which are grazed under lease arrangements. Although farmers are asked in the census to report the area of land leased, these were not captured in the geocoding exercise.

Mapping commodities

Mapping of commodities from the census data could enable the production of thematic maps, for example the location of dairy farms, as well as providing a basis for estimating internal rates of return for particular land uses within catchments.

Commodities were mapped by database matching; the dominant commodity was linked to agricultural land use (see Figure 6). Tertiary land use codes could then be developed from the secondary land use code and the ABS commodity data (Table 3).

Table 3: Development of tertiary land use codes

Agricultural activity

Secondary land use class

Dominant commodity

Tertiary land use class


3.4 Dryland permanent cropping

Non-cereals crops used for hay

3.4.3 Hay and silage


3.3 Grazing of improved pastures


3.3.2 Pasture legumes


4.5 Irrigated perennial horticulture


4.5.4 Irrigated vine fruits

After spatial matching, 41 commodities could be mapped (Table 4).

Table 4: Tertiary land uses and commodities mapped for Gippsland

Secondary land use

Tertiary land use

Examples of Commodities

2.1 Grazing of native pastures

2.1.0 Grazing of native pastures

Native pasture

3.3 Grazing of improved and fertilised pastures

3.3.1 Native/exotic pasture mosaic

3.3.3 Pasture legumes

3.3.4 Legume/grass mixtures

3.3.5 Sown grasses


Lucerne, legumes


Annual, sown

3.4 Broadacre cropping

3.4.1 Cereals

3.4.3 Hay & silage

Barley, oats, maize, millet, popcorn, rye, triticale, wheat

Hay, silage

3.5 Perennial horticulture

3.5.1 Tree fruits

3.5.4 Vine fruits

3.5.6 Flowers & bulbs


Grapes, kiwi fruit

Nurseries, flowers

3.6 Seasonal horticulture

3.6.1 Fruits

3.6.4 Vegetables & herbs


Asparagus, beans, broccoli, garlic, onions, potatoes, pumpkins, snowpeas, sweetcorn

Figure 6: Secondary and tertiary land use and commodities for part of the Gippsland study area

Mapping land management practices

DNRE funded questions for Victoria in the 1996/97 agricultural census on:

  • tillage practices – “Area of holding to which the following number of cultivations were made for fallow weed control or seedbed preparation prior to sowing broadacre crops”, reported as the number of passes prior to cultivation (0, 1-2, > 2 passes).
  • disposal of crop and pasture stubble prior to sowing – “Area of holding on which pasture and crop stubble prior to sowing pastures and broadacre crops,” reported as hectares burnt; removed by baling, heavy grazing, fire harrowing; ploughed into the soil; mulching or left intact.
  • fallow –“How much land (hectares) in fallow through the following preparations; complete chemical fallow, pasture topping or cultivation”.

Table 5 shows the level of farmer responses to these land management practice questions. The response rate is based on an estimate of the potential number of respondents, namely farmers reporting crops and/or improved pastures.

Table 5: Response to land management practices census questions

Land management practice

Area reported (ha)

Responses/Potential responses

Response rate

Tillage practices




Stubble management




Fallow treatment




The response rate for land management practices was poor, for example only 21% of farms reporting crops (33% of the cropped areas) reported tillage practices. Stubble management had the poorest response rate if all farmers who were reporting improved pastures are included as potential respondents. The response rate improves to 15% (representing about 6% of cultivated areas) if only farmers undertaking cropping are reporting stubble management activities. For the fallow treatment question, only 9.5% of farms reporting cropping provided a response, (representing less than 12% of the cultivated area).

The farms reporting these management practices were matched to the broad agricultural land uses as well to the commodity information mapped from the geocoded census data (Table 6).

Table 6: Matching land management practices to agricultural land use

Land management practice

Number of farms reporting in geocoded area

Number of farms matched to cropping land use

Area (hectares) matched

% of cropping land use

Tillage practices





Stubble management





Fallow treatment





Total reporting *





* includes one or more land management practices

Practices could be matched to cropping land use for about one third of the geocoded farms reporting land management practices. This suggests that that either the farmers have reported land management practices when in fact they were carrying out grazing or horticulture or that the cropping has not been correctly captured in the land use data. These factors plus the very poor response rate of farmers to land management practices questions resulted in only 11% of the cropping area being matched to land management practices.

Figure 7 is an example of the tillage practices matched to cereal commodities. Responses on tillage practices from a quarter of the farmers could be matched to commodity data. This provided mapped data for 1,133 hectares or 10% of the area cropped in the geocoded data.

Figure 7: Tillage management practices and crop types in Gippsland

The data mapped for land management practices are fairly sparse as a result of both poor farmer response rate and the small area cropped in Gippsland (most land management practices are associated with cropping). The confidentiality of this information could be assured by presenting the data in a grid format. For example, a grid of 5x5 km could be attributed with the area or frequency of land management practices.


Output data sets

The major outputs of this study were spatially explicit catchment scale data sets showing the distribution of commodities grown in Gippsland and some of the land management practices used by farmers. Data were presented at the land use mapping scale of 1:100,000, appropriate for mapping broadacre crops. In horticultural and irrigated areas, 1:25,000 scale land use mapping and subsequent commodity and practice mapping may be more appropriate.

The data sets, which have been created as ArcView shapefiles, can be represented in most GIS software. The data provided to ABS contain a number of attributes to aid display and querying of the data. The types of questions that can be answered using these data sets include:

  • Where are wheat or seasonal vegetables grown in Gippsland?
  • How many areas of grazing smaller than 10ha are found in the Bairnsdale statistical local area?

The geocoded census data provided spatially explicit statistical information for about half the area mapped as being under agricultural land uses by DNRE. About 75% of broadacre crops and 68% of pastures were located accurately. However, farmers’ poor response rates to land management questions resulted in practices being matched to only 37,094 hectares for secondary land use. The reasons for these results and opportunities for future improvements are presented below.

Land use mapping

According to the DNRE land use data, there are 1,452,535 hectares of agricultural land in Gippsland. The geocoded census data covered 739,880 hectares, 51% of the area mapped as agricultural land use by DNRE. The disparity between these estimates is due to the classification of certain land uses by DNRE, as well as undercoverage in the census data, incomplete geocoding and apparent under-reporting of native pastures in the census.

DNRE did not map rural residential land and these areas were probably mapped as grazing of native or improved pastures. The rural residential class (5.4.2) was established to capture “hobby farms”; farms where the EVAO is likely to be less than $5,000 and thus were excluded from the census. This would have caused an overestimate of grazing land. The extent of the overestimate is not known however delineation of rural residential properties could be made using land use planning data and local knowledge.

DNRE mapped around 202,000 hectares (5.2% of the Gippsland study area) as grazing of native pastures from departmental information on areas of crown land leased for grazing. The extent of this activity was not captured in the census or in the subsequent geocoding. Further analysis of these areas was undertaken using land tenure data. This showed that most of this grazed land was within State Forests. These may have been more appropriately mapped '2.2 Production forests', with grazing listed as a secondary use. The current land use mapping classification (Bureau of Rural Sciences, 2001) may need some modification to take account of such multiple uses.

Geocoding of farm boundaries

Boundaries for 1,069 (18%) farms on the Business Register for Gippsland were not captured in the voluntary geocoding exercise due to lack of response, a disparity between the date of geocoding and the census data or inadequate data provided by farmers. This is likely to include hobby farms, which are not currently included in the agricultural census, as well as farms not known to the ABS. The areal extent of the farms not captured by the geocoding process is estimated to be about 330,000 hectares.

Checking of the results of the geocoding exercise by ABS against local government records in two Statistical Local Areas indicated that around 20% more farms should have been included in the Business Register for these SLAs. Further checking is required to determine the extent of undercoverage across the whole of Gippsland. The proposed linking and updating of the Business Register Records with Australian Business Number information from the Australian Tax Office will help ensure that all eligible farms are incorporated in the agricultural census. These data may also help separate out “hobby farms” or rural residential properties from those where agricultural production is occurring. However, there are good arguments for including all rural land in future agricultural census data collection as land currently excluded may comprise a significant component of a region and have major impacts on natural resource outcomes.

Agricultural census data

Within the geocoded area, the area of land use that could not be matched to a commodity amounted to only 5,400 hectares or 0.8% of the area. Thus, commodity mapping was very successful. Multiple cropping or a mix of sub-dominant crops may reduce the accuracy of the attribution.

The major difficulties encountered in mapping land management practices were due to inadequate reporting of areas under some agricultural activities by farmers, plus very poor response rates to land management practice questions. Additionally, the way in which some questions on the census were phrased increased mapping problems. For example, it was not possible to disaggregate farmers’ responses on stubble management into pasture and cropping categories.

Some minor difficulties were due to differences in the dates of the census collection, geocoding of farm boundaries and the field verification of the land use data. This is especially likely for areas under crop/pasture rotation. Small features and ploughed or fallow fields may have been missed. Although the collection of agricultural land use data used satellite imagery from December 1996 to March 1997, field verification was undertaken in 1999. Over-estimates in the size of small fields, such as horticultural crops, may be due to the difficulty in identifying these on the satellite data. According to the census data, most areas of soft fruit (strawberries, gooseberries, blueberries and raspberries) are below 1 hectare or 16 pixels, this makes detection very difficult. Grapes, kiwifruit and orchards are much easier to classify having larger fields and virtually permanent canopy cover. Most vegetables have the capacity for multiple cropping, which means that field identification may be difficult when the crop areas are small or recently harvested. This could be improved by mapping horticultural areas at larger scales, such as 1:25,000.

Suggested improvements to the census include rephrasing of questions to encourage all farms to report the size of holdings, including grazing leases, accurately. At present multiple cropping and fallow land is estimated by inference. Restructuring the census questions to separate land management activities by major agricultural land uses (cropping, grazing and horticulture) would also aid in the interpretation of these data. Ways of encouraging higher response rates to land management practices questions should also be sought.

Errors in the land use data supplied by DNRE may also have contributed to difficulties in mapping commodities and land management practices. For example, incorrect identification of agricultural land uses may have occurred, broadacre crops with high vegetation indices (NDVI) may have been identified as horticulture or those with low NDVI as pastures.

Conclusions and recommendations

The availability of a catchment scale land use map plus geocoded agricultural census data enabled the mapping of commodities and certain land management practices for Gippsland. The confidentiality of the census data that contributed to these data sets is maintained by grouping the commodities prior to mapping (eg showing the distribution of a defined group of cereals rather than individual crops such as wheat, barley and oats). Further consultation with users of these data is needed to establish the most useful groupings of these commodities. The confidentiality of land management practice information will be maintained most effectively by presenting these data in a grid format. The size of the grid could be determined by the scale of the land use mapping and the need to ensure an agreed minimum number of farms are contributing to the grid. Further work is needed to establish the grid size.

The coverage and quality of the resulting data sets could be improved substantially by:

  • improving agricultural census coverage through the use of Australian Business Numbers
  • ensuring that geocoding is undertaken for all farms for which census data are collected
  • improving the response rate to land management questions
  • rephrasing census questions to improve responses to questions on the area of farm holdings, leases, the extent of multiple cropping and fallow land, and the separation of land management questions for cropping, grazing and horticultural activities.
  • ensuring that rural residential land (“hobby farms”) are identified by mapping staff and improve the level of field checking and map validation for land use mapping
  • where feasible, matching the timing of land use mapping more closely to the collection of census data
  • modifying the Australian Land Use Management Classification to account for multiple uses, such as the grazed State Forests identified in this study.

It is concluded that investment of the approximately $4.3 million required to geocode the agricultural census using the methods developed for Gippsland, would provide substantial returns through improved investment strategies for natural resources management and more efficient program evaluation. The ability to provide agricultural statistics for user defined areas of interest will increase ABS’ community, government and commercial customer base.

Based on the work undertaken for Gippsland, it is suggested that catchment scale commodity and land use management practice data sets could be produced from geocoded census data for Australia around $0.5- 0.75 million, subject to the availability of suitable land use maps and access to adequate computing resources.


Funding for the land use mapping carried out by Victorian Department of Natural Resources and Environment (DNRE), and geocoding of farm boundaries was provided by the National Land and Water Resources Audit. We wish to acknowledge the cooperation and assistance given by the staff of Australian Bureau of Statistics in Hobart and Canberra, Department of Natural Resources and Environment in Tatura and Melbourne, Victoria and Agriculture, Fisheries and Forestry - Australia.


AgStats, 1996-97. - AgStats, Small Area Agricultural Commodity Data, Australian Bureau of Statistics report 7117.0.30.001, Canberra.

Australian Bureau of Statistics (1999). Report of the geocoding survey of East and West Gippsland, Agriculture NPC, Australian Bureau of Statistics, October 1999, Canberra.

Barson, M and Lesslie, R. (2001). Land use and land management practice mapping for the Australian continent. Australian Collaborative Land Evaluation Program Newsletter Vol 12 May 2001.

Bureau of Rural Sciences (2001). Land Use Mapping at Catchment Scale - Principles and Procedures for Commonwealth-State Collaborative Land Use Mapping.

Department of Natural Resources and Environment (2000). Final report: Land use mapping of East and West Gippsland Catchment Management Authority regions (DAV30), DNRE, Tatura.

Lesslie, R., Barson, M. and Randall, L. (2000). Land Use Management Mapping for the Murray-Darling Basin, Phase 1 Report. Report to Murray-Darling Basin Commission, Landmark Project.

Previous PageTop Of Page