Mapping The Available Water Capacity Of Australian Soils

Soil Available Water Capacity (AWC) is the main source of water for vegetation and it is the potential amount of water available for atmospheric exchange. Studying its spatial distribution is crucial for agricultural planning and management and for use in biophysical modelling. The aim of this work is to obtain a continuous spatial prediction of AWC for Australia's wheat belt. We used a data set compiling 806 soil profiles to predict AWC at six depth intervals (0-5, 5-15, 15-30, 30-60, 60-100, and 100-200 cm) and different combinations of environmental information (topographic, climatic, soils, Landsat imagery, gamma-ray spectrometry) as covariates. The modelling techniques used were symbolic regression (genetic programming), Cubist (regression tree), and support vector machines. Instead of selecting a single model, we opted for using an averaging method to generate an ensemble model for a grid spacing of 500 m. We also used a downscaling technique to increase the resolution of the maps to 100 m (compliant with the GlobalSoilMap specifications) of selected sites. We observed decreasing RMSE values with the addition of extra covariates and also an expected performance decrease with depth. The downscaling technique showed satisfactory results, increasing the resolution while maintaining the model performance.