Sand Dune Vegetation Monitoring and Assessment Using UAV Remote Sensing: A Case Study for Karekare Beach Auckland Region New Zealand
There has been a rapid amount of ecosystem and biodiversity loss globally. Coastal ecosystems such as sand dunes have been particularly at risk, from climate change, residential intensification and invasive species (Haddad et al., 2015). This study aimed to analyse the effectiveness of Unmanned Aerial Vehicles (UAVs) for vegetation classification and monitoring. Two main techniques were employed for vegetation classification: traditional pixel-based and Object Orientated Image Analysis (OBIA). In recent years classification algorithms have increasing focused on OBIA because of its ability to include information not solely spectral but also shape, texture, compactness and spatial relationships (Blaschke., 2013). In this study OBIA had overall accuracies of 80.09% in the March 2019 flight compared to the next best pixel-based algorithm of 75.77% for the same March Flight. The accuracies for all classification algorithms were reduced in the November 2018 Flight, 77.61% for the OBIA algorithm. This trend of lower accuracy in November was seen in the other pixel-based classification algorithms also. While the overall accuracy was high, there were still many individual thematic classes in the study which were consistently misclassified. However, a promising result was Pampas grass/ Cortaderia selloana (an invasive species) achieved higher levels of accuracy when using OBIA classifications compared to pixel-based. UAVs represent a unique opportunity for ecological research. They are relatively inexpensive, can be launched rapidly and can access areas either to remote, dangerous and reduces the environmental impacts caused by trampling via traditional field methods. As the sensor technology, UAV platform technology and classification algorithms continue to evolve, the potential for the use of UAVs in environmental research is highly promising.