Aristotle University Of Thessaloniki (AUTH)
Εκτιμήσεις μικρής έκτασης σε ανομήλικα δασικά οικοσυστήματα, με ομαδοποίηση των δασικών διαχειριστικών μονάδων και πολυμεταβλητά Fay-Herriot μοντέλα
Abstract
dc:descriptionSustainable forest management relies on accurate estimates of forest parameters, typically obtained through sampling design. However, these designs usually provide reliable estimates only at the population level, posing challenges when estimating parameters for smaller areas, such as forest stands. The problem of estimations becomes much greater when the subjectivity of the inventory sampler comes into play. To address this, Small Area Estimation (SAE) methods, incorporating both sampling data and model-based estimators using auxiliary information, have been developed. Multivariate Fay-Herriot models leverage linear correlations between dependent variables to enhance estimation accuracy. This study focuses on applying clustering techniques to small Forest Management Units (FMUs) and employing multivariate Fay-Herriot (MFH) models to estimate key forest parameters, including stock volume, basal area, mean height, and tree diameter distribution, crucial for effective forest management. The research area encompasses the University Forest of Pertouli, utilizing data from previous management plans, past census inventories, and vegetation canopy height data. Methodologically, the research involves clustering FMUs, preprocessing data to handle outliers and identify linear correlations, and applying the area-level mixed models of Fay and Herriot. Clustering FMUs increases the sample sizes and the homogeneous areas, improving direct estimates and making possible the model-based estimates. Initially, a basic univariate Fay-Herriot (UFH) model provides reliable stock volume estimates at the aggregated FMU level. Subsequently, a temporal-like bivariate Fay-Herriot model utilized previous stock volume data to estimate current stock volume. Then, trivariate MFH model estimates stock volume, basal area, and mean height, with the most reliable results observed in MFH models that do not consider autocorrelated and/or homoscedastic random area effects. Finally, a MFH model for compositional data estimates tree diameter distribution, estimating the proportions of trees in four classes and then estimating tree frequency for three classes, contributing thus to forest structure description, with possible volume distribution and harvest removals estimation. The methodology offers reliable small area estimates of forest characteristics for the Pertouli forest, facilitating informed decision-making for effective sustainable forest management. Moreover, its applicability extends beyond this specific forest ecosystem when the modelling assumptions are met.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Georgakis, Aristeidis
- Γεωργάκης, Αριστείδης
Subjects
dc:subject × 28- Εκτίμηση για μικρές περιοχές
- Προεπεξεργασία δεδομένων
- Δειγματοληπτικό σχέδιο
- Ανάλυση σε συστάδες
- Μικτά γραμμικά μοντέλα
- EBLUP εκτιμητές
- Ανάλυση γραμμικής συσχέτισης
- Επιλογή μοντέλων/μεταβλητών
- Μεροληψία
- Δεδομένα ύψους βλάστησης/κόμης
- Δεδομένα απογραφών
- Small area estimation
- Data preprocessing
- Sampling design
- Cluster analysis
- Linear mixed models
- EBLUP estimators
- Linear correlation analysis
- Bias
- Model/variable selection
- Vegetation canopy height data
- Census data
- Γεωπονία, Δασολογία και Αλιεία
- Γεωπονικές Επιστήμες και Κτηνιατρική
- Δασολογία
- Agriculture, Forestry and Fisheries
- Agricultural and Veterinary Sciences
- Forestry
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/56095