Catchment-scale 3D mapping of depth to soil sodicity constraints through combining public and on-farm soil databases - A potential tool for on-farm management

There has been much recent effort in creating digital soil maps across large areas, but these products are rarely used by farmers and those managing the land. This is due to a variety of reasons; few maps represent soil properties of agronomic relevance, the spatial resolution and depth intervals are often too coarse, and there is limited confidence in the quality of the predicted maps. This study attempted to overcome these pitfalls by creating maps of the depth to a soil sodicity constraint across the Namoi River Catchment, a 42,000 km(2) area in eastern Australia. The maps were produced at a fine vertical support (1 cm) and horizontal resolution (30 m) using a single random forest model to a depth of 100 cm. Different exchangeable sodium percentage (ESP) thresholds of 6%, 10% and 15% were explored to define a constraint. Various publicly available spatial datasets were used as covariates, including satellite imagery, terrain attributes, climate, and geophysical data. The soil data used for modelling were collated from a national soil database, various soil surveys, as well as data collected on case study farms. The model could predict soil ESP across the catchment with a Lin's concordance correlation coefficient (LCCC) of 0.56 when tested with 10-fold cross-validation based on whole-profiles. The value of including soil data collected on-farm with regional databases was clearly demonstrated to produce maps for an individual farm, resulting in considerable improvements in model prediction accuracy. The utility of the maps for farm management was assessed by comparing with cotton and grain crop yield maps for individual paddocks on case study farms. The depth to soil sodicity constraint maps proved valuable for understanding yield variability, with constraints shallow in the soil profile generally resulting in lower yields, however, this fluctuated by crop type, and seasonal conditions. Future work should create depth-to-constraint maps for several important soil constraints, which may lead to a better understanding of variation in crop yields. An operational framework where farmers are able to add their own soil data to neighbouring and regional soil databases could have considerable benefits for broadacre agricultural industries in Australia.