Gozé, Léna
- Institutionen för skoglig resurshushållning, Sveriges lantbruksuniversitet
Sammanfattning
Environmental monitoring has become increasingly important in the current context of global ecological change. More directives and reporting guidelines are issued, hence the need for additional methods for exploiting data from environmental monitoring programmes in order to obtain relevant information about the current state of forests and landscapes. National monitoring programmes, such as the Swedish National Forest Inventory and the National Inventory of Landscapes in Sweden, are core infrastructures for describing and analysing state and change in terrestrial ecosystems. These programmes have large, but not fully exploited potential as a basis for basic and applied research. This thesis aims to develop and apply novel tools for analysing presence/absence (P/A) data from environmental monitoring programmes. Although the area of spatial statistics has been extensively studied, the issue of relating P/A data to plant abundance is an underdeveloped field that needs further attention. The primary goal of this thesis is thus to estimate plant abundance both locally and across large regions for various species. Such plant abundance estimators are derived through models for spatial distribution of plants, by using inhomogeneous point process models that are capable of modelling various categories of point patterns across the landscape, taking geographical covariate information into account. The methods are applied to data collected in the field as well as simulated data to assess the performance of the estimators of plant abundance and associated estimators of uncertainty. The results are promising and show the potential of P/A data in environmental analyses. Another objective of this thesis is to provide reliable estimators of uncertainty in different contexts, with a particular study that takes into account several sources of uncertainty when applying modelbased inference (Paper IV). That study shows that the variance of a predictor is a fairly good approximation of uncertainty in large-area surveys, whereas other components come into play when the study area is decreased.
Nyckelord
Presence/absence data; plant density; model-based inference; generalised linear models; forest inventory data; spatial point processes; uncertainty analysis
Publicerad i
Acta Universitatis Agriculturae Sueciae
2025, nummer: 2025:40
Utgivare: Swedish University of Agricultural Sciences
SLU författare
UKÄ forskningsämne
Matematisk analys
Miljövetenskap
Sannolikhetsteori och statistik
Publikationens identifierare
- DOI: https://doi.org/10.54612/a.3a4d2oki20
- ISBN: 978-91-8046-475-8
- eISBN: 978-91-8046-525-0
Permanent länk till denna sida (URI)
https://res.slu.se/id/publ/132980