Democritus University of Thrace (DUTH)
Assessment and inventory of forest parameters using medium-high resolution satellite data
Abstract
dc:descriptionMediterranean forests are characterized by high spatiotemporal heterogeneity and constitute one of the Earth’s biodiversity hotspots. The ecosystem services they provide are of great significant value and widely recognized by scientists and government institutions. However, Mediterranean forests are still being degraded due to their vulnerability to human and natural threats. Therefore, sustainable management plans have become an immediate necessity. Mediterranean countries are engaged in national forest inventories and monitoring practices, which require the retrieval of accurate forest attribute information such as tree density, basal area and grown stock volume. Due to Mediterranean forest’s heterogeneity and limited site accessibility over remote regions, earth observation data has been extremely supportive to assess forest resources. Remote sensing technology and the open access to remote sensing data provide great potential in the field of forestry and forest inventories. Moreover, the rapid progress in artificial intelligence algorithms facilitates the analysis of wide range of data. In this context, the combination of these powerful tools (remote sensing data and machine learning approaches) for the assessment of forest parameters constitutes a promising, yet challenging procedure. The present thesis examines various aspects for enhanced estimation of forest parameters, using remote sensing images and machine learning techniques. The thesis is divided into three main parts. The theoretical first part consists of four chapters. Firstly, Chapter 1 introduces the importance of Mediterranean forests, and the threats they confront (Chapter 1). Chapter 2 emphasizes the need of sustainable forest management plans and by extension forest inventory and accurate forest parameter estimation. Chapter 3 briefly presents remote sensing resources and how studies used remote sensing data to forest applications and particularly to forest parameter estimation, in Mediterranean areas. The chapter 4 is an introduction to machine learning techniques and how these approaches are applied on remote sensing and forestry field. Finally, the research questions and the objectives of the present thesis are presented.Part 2 consists of four studies, the first study (Chapter 7) is published in Remote Sensing of Environment (2017) and concerns Mediterranean forest parameters estimation using multi seasonal Landsat 8 OLI imagery and Random Forest algorithm. The second study (Chapter 8) is published in Remote Sensing Letters (2017) and concerns the relationship between growing stock volume and Sentinel-2 OLI imagery. The third study (Chapter 9) is published in International Journal of Applied Earth Observation and Geoinformation (2019) and concerns the evaluation of single-date and multi-seasonal spatial and spectral information of Sentinel-2 OLI imagery to assess growing stock volume. The fourth study (Chapter 10) is intended to be submitted for publication and consists an initial approach on meta-learning and stacked generalizations modeling for estimating grown stock volume, in Mediterranean ecosystem. Finally, Part 3 present a summary of how this thesis addresses its objectives, what were its main limitations and how this work could be enhanced with future research.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Chrysafi, Eirini
- Χρυσάφη, Ειρήνη
Subjects
dc:subject × 18- Tηλεπισκόπηση
- Δια-εποχιακές εικόνες
- Μηχανική μάθηση
- Remote sensing
- Multi seasonal imagery
- Machine learning
- Γεωπονία, Δασολογία και Αλιεία
- Γεωπονικές Επιστήμες και Κτηνιατρική
- Δασολογία
- Γεωεπιστήμες και Επιστήμες Περιβάλλοντος
- Φυσικές Επιστήμες
- Επιστήμες περιβάλλοντος
- Agriculture, Forestry and Fisheries
- Agricultural and Veterinary Sciences
- Forestry
- Earth and Related Environmental Sciences
- Natural Sciences
- Environmental Sciences
Rights
dc:rights- Statement dc:rights
-
- BY
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/48769