Data Analysis Workflow for Experiments in Sugarcane Precision Agriculture

Precision Agriculture (PA) comprises a set of tools to understand and manage inherent spatial variability within crop fields. PA relies on a variety of techniques to collect, analyze, process, and synthesize voluminous geo referenced data. However, prior to large-scale practice, PA requires a successful experimentation stage, which is the present stage of PA for the sugarcane system. This paper presents a data analysis workflow for PA experiments, including workflow application to a case study in a sugarcane area where an appreciable diversity of soil and plant attributes has been measured. Our data analysis workflow has basis on: i) removal of outliers, ii) representation of different data acquisition techniques on a common spatial grid, iii) estimation of typical "noise" level in each measured attribute, iv) spatial autocorrelation analysis for each attribute, v) correlation analysis to identify related attributes, and vi) principal component analysis to reduce the dimensionality of the attribute space. By treating the diversity of measured attributes on a common ground, the proposed analysis workflow guides further experimentation as well as selection of data acquisition technologies suitable for large-scale sugarcane PA.