3.1 Background

In Australia, the state and territory government agencies are primarily responsible for the collection and management of soils data. For the last 70 years these agencies have been collecting soil site data to meet the needs of state requirement. Soil site data and polygon map data are managed in data systems tailored to the requirements of the individual agencies. The systems were typically purpose built and supported the operating requirements of each state agency under state legislation and regulation.

Thus, in Australia, there are eight independent and unique soil data management systems running in a diverse range of software and hardware environments, but fortunately, most comply with a common semantic model as described in the Australian Soil and Land Survey Field Handbook (National Committee on Soil and Terrain, 2009).

Given this situation, it has proved difficult over the years to produce nationally consistent soils information uniformly across the continent that is readily accessible and useful at reasonably fine scale. Several attempts have occurred in the last 60 years to produce a nationally consistent soil type map.

The first attempt, the Atlas of Australian Soils (Northcote et al, 1960-68), was compiled by CSIRO in the 1960s to provide a consistent national description of Australia's soils. It comprises a series of ten maps and associated explanatory notes, compiled by K.H. Northcote and others. The maps are published at a scale of 1:2,000,000, but the original compilation was at scales from 1:250,000 to 1:500,000. Mapped units in the Atlas are soil landscapes, usually comprising several soil types. The explanatory notes include descriptions of soils landscapes and component soils. Soil classification for the Atlas is based on the Factual Key (Northcote 1979). In 1991, a digital version of the Atlas was created by the Bureau of Rural Science from scanned tracings of the published hardcopy maps. The Digital Atlas of Australian Soils is available as a shapefile. Additionally, there is a reliability map available, with a descriptive legend. The source of the reliability data is unknown (ASRIS 2018). In 2002 Ashton and McKenzie, interpreted the original Atlas data to generate for the first time a nationally consistent soil map, classified using the Australian Soil Classification (Isbell , 2002) (Figure 6.). Whilst this map and the descriptive data associated with it was useful for many purposes, it was produced at a very broad scale and mostly qualitative in nature, being generated from aerial photo interpretations with very little ground-truthing.

The second attempt at producing nationally consistent soils information was The Australian Soil Resource Information System (ASRIS) (Johnston et al. 2003). ASRIS was a collaboration of all the state and territory soil survey agencies designed to bring together all the existing soil site and polygon map data into a nationally consistent format. A range of products was developed within the ASRIS program including a “best scale” soil type map once again based on the Australian Soil Classification System (Figure 7). This polygon map was a compilation of all the existing soil mapping data across a range of scale from 1:25,000 to 1: 2,000,000 scales. Thus, in some areas the data was very accurate and in others not so accurate. The quality of the attached soil attribute data also ranged from highly quantitative to highly qualitative, and there were numerous inconsistencies across state and territory borders. As can be seen in Figure 2 the compilation map does not completely cover the entire continent.

In 2018 Teng et. al. produced the first quantitative update of a national soil map at the continental scale at a 1 km grid cell resolution (Figure 8). They produced a map of Australian Soil Classification Orders with data derived from a combination of traditional soil profile classifications and soil classifications made with visible–near infrared (vis–NIR) spectroscopy, totaling 38 756 individual observations. Digital soil class mapping (DSM) using a Random Forest model was performed. The overall error rate of the DSM model, tested on an independent validation set, was 55.6%. A detailed description of the methodology used can be found here and the raster map data can be downloaded from the CSIRO Data Access Portal.

3.2 Accuracy Comparisons

It is difficult to do a truly unbiased comparison of map accuracies for different products as previous map versions are based on expert knowledge and a large qualitative component. The Atlas of Australian Soils map also contains non mapped component information within each polygon. There is also no statistically valid external validation set available to apply to the older maps. In this work we did not have access to the independent validation data set used in the Teng et al. (2018) work, thus confirmatory accuracy assessment is not possible. However, if we take the existing observed dataset upon which this new map is based and compare the ASC classifications at these locations with those of the previous mapping efforts, we can get a general understanding for how the various maps compare.

Using this approach we calculated that the:

• Atlas of Australian Soils map has an overall accuracy of 36%,

• ASRIS "best scale" polygon map has an overall accuracy of 47%

• Teng et al. (2018) report a mapping accuracy of 55.6% using an external validation set of 12,919 observations. We calculated a map accuracy of 42% using our entire observed data set of 180,915 observations. It is possible that differences in map resolutions explain this discrepancy.

• The updated DSM modelled Australian Soil Classification map produced in this study has an overall accuracy of 61%.