Metla

National
Forest
Inventory

homepage
Executive summary
General information
History
Materials
Methods
Results
Resources
Expertise
Products
Publications
Applications of results 

Current events/projects

Use of SAR (radar) data

 Imaging spectrometry

Dynamics of biodiversity

      Finnish [Finnish]
 METLA
 METLA Research
    		
    
      


      
MULTI-SOURCE NATIONAL FOREST INVENTORY 
OF FINLAND, METHODS 

      
The Field Measurement Based Method 

      The results for large areas (200 000 hectares or over),
involving reliability assesments, are computed 
by means of field  measurements.
      

Field measurement based area estimates are point
estimates. Volume estimates of growing stock 
and increments are based on detailed measurements 
of sample trees, taper curve models and 
generalization of sample tree results to tallied trees.
      

Reliability assesments are based on quadratic forms.
The methods are desrcibed in the
publications.
      

      
The Multi-Source  Method 

      The results for small areas (about 10 000 hectares) 
are comuted by means of multi-source data. 
      

      
Satellite Image Data 

      Landsat TM 
and, to minor extent, Spot images have been applied.
The images are rectified on base maps with the pixel size of
25 m 25 m. 
      

      
Digital Map Data 

      Digital map data 
are used to separate forest and non-forest 
land from each other, to delineate computation units and
to improve the reliability of estimates. 
The applied themes  are
agricultural areas,   
built-up areas
roads,
swamp areas (peatland),
administrative boundaries
and digital elevation model.
      
                               
Digital terrain models are used in order to avoid confusion in 
image analysis caused by land morphology.
The angle between the normal to the land surface and the
sun illumination angle at the time of the satellite overpass is
used in correcting the observed spectral values. The dependence
of the normalized spectral value on the incidence angle and
the original spectral value is assumed to be inversely proportional to
a power of the cosine of the incidence angle.
      

In addition to intensity value corrections, the absolute elevation
from the sea level is also utilised in the classification phase in
such a way that a maximum distance in the elevation direction is
set from the pixel to be classified to the ground sample plots
utilised as ground truth.
      

      
Image Processing 

      The image analysis consists of preprocessing of the image
(image rectification, removal of noise, striping, etc.), selection
of features, classification, and postprocessing (generalization).
The method is chosen in such a way that all
inventory variables can be estimated for each pixel.
A k nearest neighbour classification has been applied so far.
The Euclidean distance, d, is computed in the feature space
from the pixel p to be classified to each pixel i whose ground truth is
known (sample plots).
      

Formula
      

Both areal statistics of
computation units and theme maps with land use information 
(site fertility, timber assortments and growth by tree species, etc.)
have been produced in the
classification phase.
      

The spatial information of the image can be taken into account
in the feature selection and/or in the postprocessing.
Segmentation techniques or Gibbsian random field modelling,
for example, are possible postprocessing methods.
      

      
References 

      

        
Tomppo, E. 1993. 
   Multi-Source National Forest Inventory of Finland. Proceedings of Ilvessalo
   Symposium on National Forest Inventories. Finland 17-21 August, 1992.
   IUFRO S4.02. The Finnish Forest Research Institute. Research Papers 444. pp.
   52-60.
      

      
 

      
 

      
 











VKan, December 2000