Spatial prediction for Precision Agriculture

1996

B.M.Whelan, A.B. McBratney and R.A. Viscarra Rossel
Australian Centre for Precision Agriculture, McMillan Building A05, The University of Sydney NSW 2006, Australia.

Proceedings of the 3^rd International Conference on Precision Agriculture. Bloomington/Minneapolis, Minnesota. ASA/CSSA/SSSA, pp.331-342.

ABSTRACT

Spatial prediction methods are widely used to estimate data values at unsampled locations in spatial data sets. A classification of the more common techniques is presented. The applicability and performance of these methods in the context of Precision Agriculture is outlined with reference to sample size and intensity. Local kriging with a local variogram is introduced as a means of retaining the spatial detail in intensively gathered yield data.

INTRODUCTION

A primary requirement of a Site-specific Management System is the combination of an appropriate sampling strategy with a spatial prediction method that together provide a sufficiently detailed representation of the true spatial variability of relevant crop and soil properties. At present, depending on the variable of concern, sampling intensity is dictated by convenience and the trade-off between resolution and cost. As a consequence, a variety of field information which is currently obtained by conventional sampling techniques is generally sparse in comparison with yield data generated by real-time sensors. Spatial prediction methods can be used to extend the information available from data at a sparse or uneven set of locations by estimating the values of variables at the scale of interest. Such prediction methods usually assume that the variable in question varies more-or-less continuously in space. This paper will discuss spatial prediction methods in the context of Precision Agriculture with a view to making recommendations on appropriate techniques taking into account the number and intensity of observations within a field.