Use of a Spatially Weighted Multivariate Classification of Soil Properties, Terrain and Remote Sensing Data to Form Land Management Units

Research has been conducted to develop a methodology that can delineate land management units (LMU’s) that is, zones within a paddock which can be identified, mapped and managed according to their land-use or productive capabilities. Soil sampling and analysis is a crucial component in depicting the landscape characteristics, however it is a time consuming and costly exercise to undertake. Data from a 10m resolution digital elevation model (DEM) and high resolution digital multi spectral imagery (DMSI) has been used in association with field sampled data on soil properties to investigate the variability in the landscape at large scale. The paper describes the design and implementation of a two stage methodology based on Oliver and Webster’s (1989) spatially weighted multivariate classification, for delineating LMU’s intended for precision agricultural applications. Utilising data on physical and chemical soil properties, topographic variables derived from a DEM and spectral information from DMSI collected at 250 stratified random sampling locations within a 1670 ha property in Western Australia, the methodology initially classifies sampling points into LMU’s based on a geographically weighted similarity matrix. The second stage delineates higher resolution LMU boundaries by using the geographic location, DMSI and DEM data on a 10m grid across the remaining study area and assigning each pixel to an appropriate LMU. The method groups sample points and pixels with respect to their variables and their spatial relationship on the ground, thus forming contiguous, homogenous LMU’s that can be adopted in precision agricultural applications.