The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Photonic monitoring of atmospheric fauna
Author:	Genoud, Adrien P.
View Online:	njit-etd2022-053 (xxiv, 191 pages ~ 21.6 MB pdf)
Department:	Department of Physics
Degree:	Doctor of Philosophy
Program:	Applied Physics
Document Type:	Dissertation
Advisory Committee:	Thomas, Benjamin P. (Committee chair) Wang, Haimin (Committee member) Prodan, Camelia (Committee member) Zhou, Tao (Committee member) Williams, Gregory Matthew (Committee member)
Date:	2022-12
Keywords:	Biophysics Insect Laser Machine learning Photonic sensor Remote sensing
Availability:	Unrestricted
Abstract:	Insects play a quintessential role in the Earth’s ecosystems and their recent decline in abundance and diversity is alarming. Monitoring their population is paramount to understand the causes of their decline, as well as to guide and evaluate the efficiency of conservation policies. Monitoring populations of flying insects is generally done using physical traps, but this method requires long and expensive laboratory analysis where each insect must be identified by qualified personnel. Lack of reliable data on insect populations is now considered a significant issue in the field of entomology, often referred to as a “data crisis” in the field. This doctoral work explores the potential of entomological photonic sensors to unlock some of the limitations of traditional methods. This work focuses on the development of optical instruments similar in essence to lidar systems, with the goal of counting and identifying flying insects from a distance in their natural habitat. Those systems rely on the interactions between the near-infrared laser light and insects flying through the laser beam. Each insect is characterized by retrieving its optical and morphological properties, such as wingbeat frequency, optical cross sections, or depolarization ratios. This project ran in parallel a series of laboratory and field experiments. In the laboratory, prototypes were tested and used to create a database of insects’ properties. The data were used to train machine learning classifiers aiming at identifying insects from optical signals. In the case of mosquitoes, the sex and species of an unknown specimen was predicted with a 99% and 80% accuracy respectively. It also showed that the presence of eggs within the abdomen of a female mosquito could be detected from several meters away with 87% accuracy. In the field, instruments were deployed in real-world conditions for a total of 520 days over three years. More than a million insects were observed, allowing to continuously monitor their aerial density over months with a temporal resolution down to the minute. While this approach remains very new, this work demonstrated that photonic sensors could become a powerful tool to tackle the current lack of data in the field of entomology.