Assessing and developing spatial products for agricultural applications: from field to national extent

Abstract

Identifying and understanding variability across agricultural landscapes has always been a valuable endeavour. In the modern day, these maps have been digitised into spatial products, which are represented by a grid of points, rasters or polygons associated with longitude, latitude, and a variable of interest. The potential end users of these spatial products range from growers and agronomists to industry and government. Broadacre crops such as wheat, barley, canola, and chickpea account for 35% of gross value from agriculture in Australia. The primary aim of this thesis is to assess the quality of these spatial products at the spatial resolutions required by end users within the context of agricultural cropping. This assessment is possible due to the growing volume of on-farm data now recorded and stored, specifically datasets of soil samples from agricultural landscapes, and crop yield monitor data from harvesters. Freely-accessible established and emerging spatial products relevant to four key aspects of crop production are investigated: soil, water, on-farm management, and crop productivity. The chapters of this thesis reveal key lessons for agricultural scientists and spatial modellers. Assessing global prediction quality during model testing and validation does not test the model’s ability to predict at the fine-resolution. As such, the true prediction quality of these spatial products is overestimated in accompanying literature. End users assume these values have a certain degree of accuracy. Hence, there is a responsibility for those who create these products to validate them with end users in mind, as the decisions made using information from these spatial products have national and global consequences.