University of Twente Student Theses

Login

Assessing the potential of unmanned aerial vehicle data in detecting the bark beetle infestation on European spruce

Chilongoshi, Euphrasia (2021) Assessing the potential of unmanned aerial vehicle data in detecting the bark beetle infestation on European spruce.

[img] PDF
7MB
Abstract:Insect outbreaks have caused more natural disruptions in coniferous forest ecosystems in recent decades. Early detection of bark beetle green attack, a period when trees start displaying visible signs of infestation stress, is critical to effective and timely forest management to minimize economic loss and prevent a widespread breakout. The bark beetle infestation stages include green, red, and grey attacks. The green attack is the first stage; eggs are laid beneath the tree’s bark. The second stage is the red attack; the Europen spruce tree’s needles turn from green to yellow or yellow-brown foliage. The last stage is the grey attack, which involves falling off the needles. The infestation stage identification is primarily based on field surveys, which is costly and time-consuming, especially in thick forests, as it is practically impossible to view the tree crown because of overlaps. Hence, the identification of bark beetle infestation, especially the green attack using remote sensing to complement field survey, is critical to mitigating the spread. Most studies have focused on detecting bark beetle infestation stages using remote sensing data obtained from satellites. The use of Unmanned Aerial Vehicle (UAV) offers new tools and methods for better improved early detection of bark beetle infestation by offering datasets with very high spatial resolution. The data collected by the UAV platform can be used in a variety of ways and methods to detect bark beetle infestations. The study's main primary objective was to assess the potential of unmanned aerial vehicle data in detecting the bark beetle infestation on European spruce. To obtain the multispectral UAV images, a parrot Sequoia multispectral camera mounted on a Phantom 4 UAV was flown on the three sampled blocks (one, two, and three). The field-based data collection considered the different bark beetle infestations (green, red, and grey attacks), including healthy European spruce, tree location X, Y coordinates, and block number. Tree crowns of bark beetle infestations collected in the field were manually delineated to be used as spectral signatures to classify the UAV multispectral images based on the three classifiers Random Forest (R.F), Support Vector Machine (SVM), and Maximum Likelihood (ML). Further, this study also used object-based image analysis (OBIA) to ensure that the European spruce tree crowns were divided into non-overlapping categories for accurate results. In addition, the acquired UAV multispectral images were also used to calculate the three vegetation indices, which include Green Normalized Difference Vegetation Index (GNDVI), Normalized Difference Red Edge (NDRE), and Normalized Difference Vegetation Index (NDVI). The GNDVI, NDRE, and NDVI were further used as input variables in the Species Distribution Model (SDM) to determine the performance of the vegetation indices in predicting the probability of the bark beetle infestation stages, especially the green attack. The classifiers and vegetation indices' performance was assessed to identify which classifier and vegetation index gives the higher accuracy. The findings of this study showed that a UAV multispectral image could be utilized to detect and predict the probability of bark beetle infestations stages on European spruce, especially the green attack, with the possibility of giving higher accuracies. The research found out that among the three classifiers (R.F, SVM, and ML), ML outperformed R.F and SVM. Furthermore, among the three vegetation indices, NDRE outperformed GNDVI and NDVI. Additionally, the results of this study can be extrapolated to other areas because they are promising as they show that the UAV multispectral images have potential in detecting and predicting the probability of the bark beetle infestation stages, especially the green attack.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/88777
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page