Robust and diverse multidimensional statistical moments in dual-band entomological lidar for improved real time insect monitoring

As some insect groups are declining at alarming rates, accurate and automated insect monitoring is needed to prioritize habitats for conservation. Dual-band entomological Scheimpflug-Lidar technique is a promising candidate method for real time insect monitoring: it allows the detection of thousands of flying insects per day at high temporal and spatial resolutions. The signals contain a plethora of properties which can be assigned to flight headings- and species-specific clues which may improve classification. We introduce a systematic approach to robust dimensionality reduction of entomological lidar range-time intensity matrices (time and range, 2D) of observations, into time dependent vectors (1D), and scalar values (0D) which encode features related to the flight headings-, and species-characteristics. Using this single-night dataset as a case study, we show that dual-band parameters not only confirm expected patterns of average insect melanization but also enable exploration of signal diversity such as insects that display distinct spectral signatures.