Use this URL to cite or link to this record in EThOS:
Title: Remote sensing methods for biodiversity monitoring with emphasis on vegetation height estimation and habitat classification
Author: Petrou, Zisis
ISNI:       0000 0004 5365 995X
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
Access from Institution:
Biodiversity is a principal factor for ecosystem stability and functioning, and the need for its protection has been identified as imperative globally. Remote sensing can contribute to timely and accurate monitoring of various elements related to biodiversity, but knowledge gap with user communities hinders its widespread operational use. This study advances biodiversity monitoring through earth observation data by initially identifying, reviewing, and proposing state-of-the-art remote sensing methods which can be used for the extraction of a number of widely adopted indicators of global biodiversity assessment. Then, a cost and resource effective approach is proposed for vegetation height estimation, using satellite imagery from very high resolution passive sensors. A number of texture features are extracted, based on local variance, entropy, and local binary patterns, and processed through several data processing, dimensionality reduction, and classification techniques. The approach manages to discriminate six vegetation height categories, useful for ecological studies, with accuracies over 90%. Thus, it offers an effective approach for landscape analysis, and habitat and land use monitoring, extending previous approaches as far as the range of height and vegetation species, synergies of multi-date imagery, data processing, and resource economy are regarded. Finally, two approaches are introduced to advance the state of the art in habitat classification using remote sensing data and pre-existing land cover information. The first proposes a methodology to express land cover information as numerical features and a supervised classification framework, automating the previous labour- and time-consuming rule-based approach used as reference. The second advances the state of the art incorporating Dempster-Shafer evidential theory and fuzzy sets, and proves successful in handling uncertainties from missing data or vague rules and offering wide user defined parameterization potential. Both approaches outperform the reference study in classification accuracy, proving promising for biodiversity monitoring, ecosystem preservation, and sustainability management tasks.
Supervisor: Stathaki, Tania Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral