Integrating multi-source remote sensing data for soil mapping in Victoria

Mapping of soils at paddock level is being undertaken at 3 selected sites in Victoria by using remote sensing data from a variety of sources. The three sites represent different agricultural practices in high and low rainfall zones across Victoria. The project work is divided into 4 stages: (1) soil analysis at sample points, (2) analysis and integration of remotely sensed information, (3) exploring relationships between soil properties and remotely sensed variables, and (4) extrapolating soil mapping to a wider area using relationships established at stage 3. Detailed soil surveys were initially conducted to characterise the soils of each site at the sub-paddock scale to obtain information on the physical and chemical properties of the soil, particularly those relating to limiting factors such as salinity, toxicity (boron, aluminium), acidity, alkalinity and soil moisture regimes. To achieve this, exploratory soil cores were collected across the site from locations representing different soil types and landform patterns. Soil pits (ranging between 4-10 per site) were then dug. Chemical and physical analyses were also completed for each pit to assist in understanding the soil characteristics across each site and down the soil profile. This site specific soil information is then used to aid soil feature extraction from the remotely sensed data sets. High resolution remote sensing information acquired for this study includes airborne synthetic aperture radar (PACRIM 2 AIRSAR), Landsat 7 ETM+, conductivity measurements from EM38 and EM31, and ground penetrating radar data sets. All data sets are integrated into raster format. The sets of remotely sensed data are treated as explanatory variables with soil properties as response variables. The integrated approach enables a number of space, air and ground borne data sets to be evaluated in terms of their potential to be able to identify soil components both on the surface and at depth. The differences in the resolution of these sensors also allow evaluation on the type of soil information that can be obtained at varying scales, from the sub-paddock to catchment