Multi-thematic delineation of ‘natural zones’ of arable fields and their correspondence to spatial yield variation

Abstract

Properties such as soil apparent electric conductivity (ECa), topography and other site-related data (e.g. canopy reflectance from aerial images) vary across field. The agronomic effects of such variability can sometimes be seen in the spatial variations of crop yield on that field. However, yield maps do not always represent the natural boundaries based on site characteristics. Identification of these boundaries as “management zones” (MZ) can be beneficial in crop management and improving crop input use efficiency. A simple methodology is required to delineate such zones. This research presents an effective methodology to delineate MZ in an irrigated and a non-irrigated (rain-fed) arable maize field in New Zealand Elevation data for the sites were acquired from Google Earth images and a soil survey. Soil ECa was collected from a soil survey with an electromagnetic device. Yield values (t/ha) were obtained from combine harvesters equipped with yield monitor and Global Positioning System (GPS), over the course of four years for the irrigated site, and two years for the non-irrigated site. The yield data was quality controlled using a filtering system to remove outliers and technically non-plausible data. The data sources were combined in Geographic Information Systems (GIS) and three MZ were delineated for each field through standard clustering methods. The maize yields were aggregated per derived MZ to compare yields between different MZ-classes. The results showed that there was some consistency in yields related to the MZ, derived without yield data. In both the non-irrigated and irrigated fields, the lowest yield consistently occurred in the same class each year, however, the MZ-class with the highest yield varied year to year. The results show that it is possible for the studied type of fields to delineate ‘natural’ clusters or zones of site properties that can be used as MZ-classes as they represent different yield levels. The required inputs are freely available and easily obtained data.