Identifying the spatial variability of soil constraints using multi-year remote sensing

In north-eastern Australia, soil attributes such as salinity, sodicity, acidity, and phytotoxic concentrations of chloride constrain the growth of crops. It is difficult to delineate constrained areas using conventional sampling methods. Alternatively, where crops fail, over multiple years, to pass a certain yield threshold, we might infer the presence of a soil constraint. For a wheat-growing farm over 10-year period we used remote sensing to obtain a large volume of surrogate yield data, by calibrating an archive of Normalised Difference Vegetation Index (NDVI) to an archive of (limited) ground-based observations. The model used was a generalised additive model that related wheat yield as a non-linear function of NDVI and a linear function of post-anthesis rainfall. Field locations where predicted yield consistently failed to reach the 75th percentile in a given year, over a number of years, we regarded as limited by at least one unknown soil constraint. Soil samples averaged for the constrained locations showed, compared with the unconstrained locations, relatively high concentrations of subsoil chloride, and, in the topsoil, relatively high exchangeable sodium percentage, and unused nitrate nitrogen. On-farm experiments suggested that, for constrained areas, preclusion of monoammonium phosphate (MAP) fertiliser application, coupled to gypsum amelioration, could potentially benefit the farm by A$32/ha/year (MAP) and A$207/ha/3 years (gypsum).