Proximal remote sensing to differentiate nonviruliferous and viruliferous insect vectors – proof of concept and importance of input data robustness

Abstract

Proximal remote sensing is being widely studied as a noninvasive method to partially automate diagnostics of plants and insects. The hypothesis that proximal remote sensing can be used to differentiate specimens of adult beet leafhoppers (Circulifer tenellus) that were nonviruliferous or viruliferous for beet curly top virus (BCTV) was tested. A key aspect of applications of proximal remote sensing is the ‘robustness’ or repeatability of input reflectance data. Many factors may contribute to low input reflectance data robustness; these include: (i) issues related to the consistency of proximal remote sensing conditions (light intensity and spectral composition, ambient temperature), (ii) insect specimen preparation (projection angle, storage and handling), and (iii) insect specimen characteristics (age, growing conditions, variety/biotype, host plant). This study demonstrates that nonviruliferous and viruliferous specimens of adult beet leafhoppers possess unique body reflectance features and, therefore, can be differentiated. However, insect specimen preparation (removal of wings and placement) markedly affected the classification accuracy. Addition of experimental noise to input reflectance data was conducted to simulate varying degrees of input reflectance data robustness. The potential of developing reflectance‐based diagnostic tools for detection of plant pathogenic viruses in insects is discussed, with an emphasis on input data robustness.