Insect Diversity Estimation in Entomological Lidar
(2022) BINP52 20212Degree Projects in Bioinformatics
- Abstract
- The measurement of biodiversity has been a challenging topic in recent years. Our group uses a remote sensing technique named Scheimflug lidar to monitor insect activity and abundance. A field campaign was conducted back in 2020 at Stensoffa, Sweden, with our lidar system to measure insect activities over a lake. This thesis work aims to develop a method to estimate insect diversity of the data recorded from the Stensoffa campaign. Hierarchical clustering was applied to the data set to group the observations based on both waveforms recorded by the lidar or modulation power spectra obtained from waveforms. To estimate the number of clusters, we proposed an analytical model to fit an excess linkage based on the linkages of signals and... (More)
- The measurement of biodiversity has been a challenging topic in recent years. Our group uses a remote sensing technique named Scheimflug lidar to monitor insect activity and abundance. A field campaign was conducted back in 2020 at Stensoffa, Sweden, with our lidar system to measure insect activities over a lake. This thesis work aims to develop a method to estimate insect diversity of the data recorded from the Stensoffa campaign. Hierarchical clustering was applied to the data set to group the observations based on both waveforms recorded by the lidar or modulation power spectra obtained from waveforms. To estimate the number of clusters, we proposed an analytical model to fit an excess linkage based on the linkages of signals and noises. We also compared this method to the L method that found the knee point of the curve. To investigate the influence of the instrument complexity, we evaluated the benefit of polarization band features of the instrument to improve specificity. Furthermore, we applied the analytical model to different time intervals to investigate the relationship between the number of observations and the estimated number of clusters. In addition, we developed an approach to measure pairwise similarity for the waveforms and compared this to the phase insensitive modulation power spectra. (Less)
- Popular Abstract
- Insect Diversity Estimation
Global biodiversity has been suffering a significant loss in recent years. It is essential to develop a method to measure biodiversity. In this work, we used a lidar (laser radar) system to monitor insect activity, and I developed a method to estimate how many types of insects the data set contained.
A common method to measure insect diversity is using traps to attract insects and identify each caught insect, which is expensive and time-consuming. Our lidar system can monitor insect activities in a broader range without bothering them. The insect activities were recorded as waveforms with shapes similar to analog signals, and we compared the similarity between waveforms and their modulation power to... (More) - Insect Diversity Estimation
Global biodiversity has been suffering a significant loss in recent years. It is essential to develop a method to measure biodiversity. In this work, we used a lidar (laser radar) system to monitor insect activity, and I developed a method to estimate how many types of insects the data set contained.
A common method to measure insect diversity is using traps to attract insects and identify each caught insect, which is expensive and time-consuming. Our lidar system can monitor insect activities in a broader range without bothering them. The insect activities were recorded as waveforms with shapes similar to analog signals, and we compared the similarity between waveforms and their modulation power to estimate how many clusters the data set contained. Our data set had 94,039 insect signals, and we aimed to estimate how many types of insects were in the data set.
To group the insect signals into clusters, I applied hierarchical clustering to the data set to generate a tree, as shown in Fig 1.
Each leaf node represents a signal and a cluster itself. The most similar clusters will merge up until there is only one cluster. My task is to decide where to stop merging clusters and calculate how many clusters are left, represented by the dashed line in Fig 1. I also investigated the factors that could influence the estimation, including signal length, modulation, phase information, instrument features (polarization), and the size of the data set.
Through my results, I proposed an analytical model for the result of hierarchical clustering. The best estimation of the number of clusters was 55, which means there were at least 55 types of insects in the data set. In addition, I developed an algorithm to calculate the pairwise similarity of waveforms by shifting one waveform within another waveform relatively in time. The results also showed that insects were especially active around sunrise and sunset, and insect activities increased as the temperature decreased.
In this work, I applied hierarchical clustering on waveforms, and the results showed that phase information was helpful for clustering. The proposed analytical model was robust to estimate the number of clusters for different modulation and polarization bands. Signals with long lengths provided a better estimation of the number of clusters. As insects were especially active around sunrise and sunset, conducting field campaigns to observe or catch insects would be more efficient within these periods and temperatures.
Master’s Degree Project in Bioinformatics 60 credits 2022
Department of Biology, Lund University
Advisor: Mikkel Brydegaard
Advisor’s Department: Department of Physics, Lund University (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9102879
- author
- Xu, Zhicheng
- supervisor
- organization
- course
- BINP52 20212
- year
- 2022
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 9102879
- date added to LUP
- 2022-11-07 11:34:08
- date last changed
- 2022-11-07 11:34:08
@misc{9102879,
abstract = {{The measurement of biodiversity has been a challenging topic in recent years. Our group uses a remote sensing technique named Scheimflug lidar to monitor insect activity and abundance. A field campaign was conducted back in 2020 at Stensoffa, Sweden, with our lidar system to measure insect activities over a lake. This thesis work aims to develop a method to estimate insect diversity of the data recorded from the Stensoffa campaign. Hierarchical clustering was applied to the data set to group the observations based on both waveforms recorded by the lidar or modulation power spectra obtained from waveforms. To estimate the number of clusters, we proposed an analytical model to fit an excess linkage based on the linkages of signals and noises. We also compared this method to the L method that found the knee point of the curve. To investigate the influence of the instrument complexity, we evaluated the benefit of polarization band features of the instrument to improve specificity. Furthermore, we applied the analytical model to different time intervals to investigate the relationship between the number of observations and the estimated number of clusters. In addition, we developed an approach to measure pairwise similarity for the waveforms and compared this to the phase insensitive modulation power spectra.}},
author = {{Xu, Zhicheng}},
language = {{eng}},
note = {{Student Paper}},
title = {{Insect Diversity Estimation in Entomological Lidar}},
year = {{2022}},
}