Note: This dataset description is metadata (data about data) which describes the actual dataset in accordance with the ANZLIC (Australia New Zealand Land Information Council) Core Metadata Guidelines
ANZLIC unique identifier: ANZCW1202000134
Title: SOIL PH FOR AUSTRALIAN AREAS OF INTENSIVE AGRICULTURE OF TOP SOIL (DERIVED FROM SITE MEASUREMENTS)
Custodian: CSIRO, Land and Water
Jurisdiction: Commonwealth
Abstract:
This dataset presents a surface of predicted pH (X 1000) of layer 1 (A Horizon - top soil) surface for the intensive agricultural areas of Australia. Data modelled from spot observations taken by soil agencies both State and CSIRO. For details of modelling methodology, see Henderson et al 2001. The digital map data is provided in geographical coordinates based on the World Geodetic System 1984 (WGS84) datum. This raster data set has a grid resolution of 0.001 degrees (approximately equivalent to 1.1 km). The distribution of soil pH can be used to assess soil acidity, a chemical condition of the soil that reduces crop and pasture yields. It is known as surface acidity when it is in the topsoil or ploughed layer (roughly 0-30 cm depth). The data set is a product of the National Land and Water Resources Audit (NLWRA) as a base dataset. An additional point model based pH layer 1 prediction surface for South Australia is available. Uncertainty surfaces corresponding to ASRIS point model prediction surfaces have been prepared for this soil property. The dimensionless uncertainty values fall in the range 0 to 1000 where lower values represent greater uncertainty.
ANZLIC search words:
Spatial domain:
Beginning date: 1999-09-01
Ending date: 2001-03-31
Progress:
Maintenance and update frequency:
Access constraints:
license
Lineage:
pH in water (method 4A1) was standardized to pH in calcium chloride (Australian Soil and Land Survey Laboratory Handbook methods 4B1,4B2). The conversion equation used was that of Ahern (1995) pH_Ca = 0.93*pH_w - 0.373. Measurements corresponding to other pH methods or ratios other than 1:5 were ignored. Modelling --------- The models were created using all the standardized pH data simultaneously, i.e. single Australia-wide models were developed. These were based on 11051 measurements of layer 1 pH The models were created using the regression tree software Cubist (see www.rulequest.com) and a suite of 40 environmental predictors (19 climatic surfaces,4 bands of MSS, lithology, DEM and 15 DEM-derived predictors). The layer 1 pH model uses Cubist options that generate a "committee" model where 3 tree models are developed, each tree targeted slightly differently, and an average of the 3 predictions used to give the final prediction. The null space that exists in the prediction surfaces represents areas where predictions could not be made because the combinations of environmental predictors in those areas do not satisfy any of the constraints in the rules. While these areas are invariably poorly represented in terms of samples, it is considered that this effect is largely due to the coarseness of the climatic surfaces. Moreover, it is thought that this coarseness is responsible for some of the banding that is evident in the predicted surfaces. The residual plots highlight the sparsity of sampling in some areas in the ASRIS extent. There is less confidence in the predictions in poorly supported areas. At this stage the point density is the only surrogate for model uncertainty provided. The dataset was resampled, increasing the cell size to 0.01 deg, or approximately 1.1 km using bilinear interpolation. Uncertainty surfaces corresponding to point model based ASRIS prediction surfaces were prepared. The dimensionless uncertainty values fall in the r
Positional accuracy:
Input sample locations have variable positional accuracy most locations are expected to be within 100m of the recorded location. Where vertical information is available it is expected to be within 20 mm of the recorded depth below surface. No assessment of vertical position to the GDA datum has been attempted as it is not considered relevant. The positional assessment has been made by considering the tools used to generate the locational information and contacting the data providers. The other parameters used in the production of the modelled surface have a range of positional accuracy ranging from + - 50 m to + - kilometres. This contributes to the loss of attribute accuracy in the modelled surface. The modelled predictive surface is a 250m grid and has a locational accurate of about 1m. To provide consistency the dataset was resampled, increasing the cell size to 0.01 deg, the positional accuracy has not been altered in this process.
Attribute accuracy:
All pH values have been multiplied by 1000 to enable the use of integers (divide value by 1000 to give appropriate pH at site) The values in the table overemphasis the real accuracy of the data. Input sample attribute accuracy was provided at one decimal point and is deemed to be + - .1 pH. The predictive values has a variable and much lower attribute accuracy due to the irregular distribution and the limited positional accuracy of the parameters used for ling. Predicted average soil pH X1000 and classed average soil pH in the 0.01 X 0.01 degree quadrat VALUE (Binary Integer), CLASS_VALUE (Binary Integer) and CLASSES (Character) VALUE CLASS_VALUE CLASSES pH * 1000 pH Class lt = 4300 1 lt = 4.3 gt 4301 - gt = 4800 2 gt 4.31 - lt = 4.80 gt 4801 - lt = 5500 3 gt 4.81 - lt = 5.50 gt 5501 - lt = 7000 4 gt 5.51 - lt = 7.00 gt 7001 - lt = 8500 5 gt 7.01 - lt = 8.50 gt 8500 6 gt 8.50 Uncertainty surfaces: a dimensionless uncertainty number (Field =Value) falls in the range 0 to 1000 where lower values represent greater uncertainty.
Logical Consistency:
Surface is fully logically consistent as only one parameter is shown, as predicted average pH X 1000 within each grid cell.
Completeness:
Surface is nearly complete. There are some areas (about %2 missing) for which insufficient parameters were known to provide a useful prediction and thus attributes are absent in these areas.
Metadata date:
This dataset description can be discovered via the Australian Spatial Data Directory (ASDD)