Repository logo
 

Monitoring New Zealand's Native Bees: A Collaborative Approach Using Image Analysis

Files

Whole thesis(27.84 MB)

Date

2016

Authors

Supervisor

Huang, Loulin

Item type

Thesis

Degree name

Doctor of Philosophy

Journal Title

Journal ISSN

Volume Title

Publisher

Auckland University of Technology

Abstract

New Zealand has around thirty different species of native bees. They are pollinators of wild and cultivated plants, and are likely important to the health of ecosystems. Most species are solitary ground nesting bees; individual females construct their nests in the ground. During the active flight season, thousands of bees will nest alongside each other to form large communities. Although much is known about the biology of native bees, studies can be difficult. As a result, there is much to learn about their diversity and population status. To address this problem, a method to measure populations of native bees using digital images and analysis was proposed. While it was difficult to acquire images of individual bees, it was straightforward to photograph nests. Furthermore, the number of nests could provide an indicator of community health. For these reasons, the methods in this research focused on counting the number of active nests. Surveys were conducted over six years (2009--2014), at three communities of native bees located in Whangarei (Northland, New Zealand). Monitoring data were collected across five years (2010--2014). Fundamental ecological data were collected, including manual nest counts; digital images of active nests were acquired. Open source, biomedical imaging platform {FIJI}, was used to process images. Nest counts derived from the automated imaging methods were compared with manual-field and manual-image counts. There were good agreements between methods. Results suggested image-centric monitoring methods could replace manual-field nest counting methods. Data by manual and automatic techniques, indicated the number of active nests have decreased over five years. The imaging methodology presented in this thesis shows good potential. The image-centric design was fully documented, based on open source software and off-the-shelf tools. Therefore, the system could be immediately adapted for other environments and provide the tools to gather much needed information about the health of important background pollinators throughout New Zealand and the world.

Description

Keywords

Image analysis, Random forest, Native bees, Solitary bees, Community science, Ecological monitoring

Source

DOI

Publisher's version

Rights statement

Permanent link

https://hdl.handle.net/10292/9617

Collections

Doctoral Theses