Applying advanced geospatial technologies to capture variation in soil health variables in hill country landscapes

This study aims to quantify the spatial variation in soil health variables in a complex agroecological landscape using modern geospatial analysis tools and technologies. A wide range of soil attributes (soil organic carbon concentration, soil nutrient fertility, bulk density, and earthworm abundance), underlying factors describing the topographical characteristics of the land surface biophysical pattern, along with land use management practices, were utilised to model the spatially explicit pattern of each soil health component. Machine learning techniques were applied to predict soil attributes at a pixel-level across the whole landscape from a limited number of soil samples collected from specific locations. Soil health was quantified using a Composite Soil Health Index (CSHI), calculated from the mean value of the standardized individual soil health indicator which is obtained from the scoring functions for each grid cell. The approach was applied across farmlets that make up the long-term phosphorus (P) fertiliser and sheep grazing experiment at Ballantrae located near Woodville (Southern Hawke’s Bay, New Zealand). Results from our study reveal that the variables contributing to soil health varied both across the landscape and between soil health indicators. The study demonstrates that advanced spatial statistical analytics and remote sensing can be effective tools to address the challenge posed by the modelling of biophysical processes in complex agroecological landscapes. Applying such an approach provides a more complete picture on soil health and therefore, can advance the environmental planning and management of farms in New Zealand.

NZBIDA