|
|
My research interests
Mixed-effects models
MIxed-effects models are a natural way to model
such grouped data sets where the goups are a sample from a larger
sample of groups. The modeling framework provides a convenient framework
both for inference and prediction using such data. About 50% of
my publications apply mixed-effects models in various fields, including
forest mensuration, ecology, remote sensing, and game research.
The models include linear mixed-effects models with various kind
of grouping structures, nonlinear mixed-effects models, generalized
linear mixed-effects models and systems of mixed-effects models.
Remote sensing
My earlier research and PhD was about modelling
forest stand structure, and especially tree size (=diameter distribution)
and basic allometric relationships among trees of a stand (H-D curve).
Together with some additional models (that for spatial pattern of
tree locations and allometric relationships between tree size and
crown properties), they essentially define probabilistic properties
of a three-dimensional object called forest stand. In remote sensing,
we are taking either two or three-dimensional observations from
this object. Remote sensing takes observations from a realization
of such model, and aerial forest inventory tries to infer forest
parameters from such data.
Forest biometrics
The traditional topics in forest biometrics are
models for site and stocking, modelling of tree and stand growth,
modelling ingrowth and mortality, allometric relationships of trees,
taper curves and volume models, and diameter distributions. I have
conducted and I am currently conducting some research on these topics.
My special interest is utilizing of sample measurements to improve
the predictions from these models, by using the best linear (unbiased)
predictor (BLP or BLUP). This terminology is closely related to
mixed-effects models, but can be generalized also to other models.
Diameter distributions
With diameter distributions, my main interests are
(i) how to take into account
the mathematical relationships of stand variables and the diameter
distribution into account in modelling, (ii) how to improve the
prediction by using sample information, and (iii) modelling of percentile-based
diameter distribution.
Forest planning
My research on forest palnning is related to cost-efficient
use of the forest data in planning. In addition, I supervise theses
on the effects of different sources of inaccuracy to the quality
of a forest plan.
Current collaborators
National
- University of Jyväskylä (Department
of statistics),
- University of Helsinki (Department of forest
sciences),
- University of Eastern Finland (All faculties)
- University of Oulu (Department of biology)
- Natural Resources Institute Finland (e.g. groups
in NFI, statistics, game research, ...)
International
- Yale University (CT, USA)
- Virginia Tech (VA, USA)
- University of Natural Resources and Life Sciences,
Vienna, Austria
- Brazil (Federal Technological University of Paraná
and University of Sao Paulo)
- I also have some more or less active collaboration
with several other national and international organizations.
Return to main page |
|