Use of Unfold PCA for on-line plant stress monitoring and sensor failure detection

Abstract

Phytodiagnostic tools are aimed at the early detection of plant and/or crop stress conditions, with a view to assisting farmers to timely intervene in order to prevent or limit plant damage. Continuous monitoring of the variations in stem diameter has attracted a lot of attention in the assessment of the plant's response to stress conditions, but on-line data acquisition leads to large datasets. Unfold Principal Component Analysis (UPCA) is a powerful and conventional tool for intelligent handling of large datasets and on-line process monitoring of batch processes in the absence of exact process knowledge. Although the use of UPCA is already common practice for certain biotechnological applications, practical applications in plant science with respect to on-line plant stress monitoring are still lacking. This work describes for the first time the application of this technique to datasets of two different plant species (i.e. young apple trees and truss tomato plants) with a view to the development of an early warning system for stress detection in plant crops. To this end, UPCA modelling was applied in such a way that a single diurnal cycle corresponds to a single batch in the original methodology. For both plant species, a PCA model was calibrated using the data of the first days of the experiment of a fault-free control plant. The remaining data of the control plant were projected onto this model for validation. Projection of experimental data of a stressed plant onto the same PCA model allowed successful stress detection, days before the appearance of visible symptoms. It can be concluded that successful alarm generation is possible with UPCA.