Spatial and temporal prediction of soil properties from legacy data

Soil monitoring networks are expensive to implement and can only predict rates of change once soil properties have been measured twice. Legacy soil datasets can act as cheaper substitutes and can yield immediate assessments of change but problems can arise because they are not designed for this purpose. We demonstrate that geostatistical techniques can predict spatial and temporal trends in a dataset of soil samples submitted by farmers across Victoria, Australia. Expert interpretation is required to determine whether observed trends correspond to changes in underlying soil properties rather than changes in farmers' priorities when requesting soil tests. The estimated models of variation can also be used to design soil monitoring networks although expert interpretation is again required to assess the extent to which the variation of the legacy data reflects that from a purpose-built survey.