Author

Nial J. Roder

Year

2017

Degree Name

BEnvSc Hons

Department

School of Earth & Environmental Science

Advisor(s)

Sarah Hamylton

Abstract

Anthropogenic influences have disturbed the natural functions and characteristics of ecosystems worldwide, and the restoration of such ecosystems is a common challenge faced by environmental managers. The Five Islands Group, Port Kembla, south of Wollongong City, has a long history of human interaction which has caused significant changes to the vegetation communities on the islands. The Five Islands were listed as a nature reserve in 1960 as they provide an important breeding habitat for many species of native seabirds. The largest of the islands, known as Big Island (17.7 ha), has shifted from an ecosystem supporting mainly native vegetation communities to one that is dominated by the exotic species Kikuyu grass and Coastal Morning Glory, resulting in severe habitat degradation for native seabirds that breed on the island. This project uses remote sensing and GIS techniques to map the distribution of native and exotic species on Big Island. High spatial resolution images were acquired using an unmanned aerial vehicle (UAV), which has emerged as a popular platform for remote sensing of the environment in recent years. Land cover classification maps were produced using a supervised maximum likelihood classification. As part of an ongoing seabird habitat restoration project, New South Wales National Parks and Wildlife Service conducted an aerial helicopter spray on Big Island in April 2017 to eradicate weeds. Thematic land cover maps were produced from images acquired before and after the aerial treatment.

A validation exercise was conducted and involved the collection of an independent georeferenced photographic dataset of the ground cover on Big Island to compare against the thematic content of the maps. The validation exercise yielded overall accuracies of 84% for pre-treatment classification and 75% for the post-treatment classification. The land cover maps provide the first detailed, high resolution pixel-based classification maps of Big Island, which can be used to establish a baseline against which future changes can be monitored. Land cover classes within the treated area before and after treatment were compared to assess the effectiveness of the spray in destroying the target species. The analysis revealed Kikuyu within the treatment area accounted for 66% of land cover before the treatment, and was reduced to 1% of land cover following treatment. Percentage cover of Coastal Morning Glory was reduced from 15% to 1% following treatment. Additionally, a series of historical aerial photographs were examined to estimate the trends of vegetation cover on Big Island. The aerial photographs revealed a rapid increase in vegetation cover following the introduction of Kikuyu grass to the island. Four additional permanent vegetation survey plots were established to supplement seven other previously established plots. A 20 × 20m quadrat was sampled at each plot location, following the NSW vegetation survey standards set out in the Native Vegetation Interim Type Standard (Sivertsen, 2009). All four sampled quadrats were dominated by exotics.

Vegetation mapping using remote sensing and GIS techniques has been recognised as a useful tool for monitoring the environment, and its applications will continue to improve with the development of higher quality spatial and spectral resolution sensors (such as UAVs). However, this should not deter from the benefits of field visits in understanding an ecosystem of interest.

FoR codes (2008)

050205 Environmental Management, 050206 Environmental Monitoring

Share

COinS
 

Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.