vay to 
d and 
DEM 
Mean 
1S€. 
to our 
ROM 
80 
metry 
1g 2D 
metry 
SING 
Juha Hyyppä 
AUTOMATIC DERIVATION OF FEATURES RELATED TO FOREST STAND ATTRIBUTES USING 
LASER SCANNER 
Juha HYYPPÁ', Hannu HYYPPÄ”, Georg RUPPERT““ 
"Finnish Geodetic Institute, Finland 
Department of Photogrammetry and Remote Sensing 
Juha.Hyyppa@fgi.fi 
"Helsinki University of Technology, Finland 
Department of Electrical and Communications Engineering 
Hannu.Hyyppa@hut.fi 
""Joanneum Research, Austria 
Institute for Digital Image Processing 
Georg.Ruppert@joanneum.ac.at 
KEY WORDS: Laser scanning, segmentation, watershed, forest inventory, forest stand attributes. 
ABSTRACT 
High-pulse-rate laser scanners are capable to detect single trees in boreal forest zone, since significant amount of laser 
pulses reflect directly from the ground without any interaction with the canopy. This allows detailed investigation of 
forest areas and the creation of a 3-dimensional tree height model. By extracting the height, location and crown 
dimension of the trees from the 3-dimensional tree height model obtained from laser scanner data, important tree 
attributes, such as stem volume, basal area, and age, can be estimated for single trees. By knowing the characteristics of 
single trees, forest characteristics for sample plots, stands and larger areas, such as stem volume per hectare [m’/ha], 
basal area per hectare [m*/ha], mean height, dominant height, mean age, number of stems [pc/ha] and development class, 
can be calculated. The automatic extraction of crown dimensions from 3-dimensional tree height models is based on 
segmentation. This paper describes the general method and gives a first indication of the performance of the developed 
method. The following standard errors were obtained for mean height, basal area and stem volume at stand level: 2.3 m 
(13.6 96), 1.9 m*ha (9.6 %), and 16.5 m’/ha (9.5 %), respectively. The accuracy was better than the accuracy of 
conventional standwise field inventory. 
1 INTRODUCTION 
One of the most potential applications using high-pulse-rate laser scanners is the forest inventory at the stand level. 
Forest stand is the smallest description unit of forests and it represents a homogeneous forest area with respect to forest 
resources and treatments needed. The size of the forest stand varies, normally, between 0.5 and 5 hectares. Presently, 
forest inventory data is collected primarily by means of field surveys, which is both expensive and time-consuming. 
Important forest attributes, including stem volume per hectare, are then assessed to these stands by measuring sample 
plots and individual trees, and by using personal experience. At this level of planning, the accuracy requirements on 
forest data are very high, typically a 15 % error is tolerated (standard error divided by the mean). Typical costs per 
hectare range between 15 and 20 USD/Euro of which about 50-60 % is used for data acquisition and recording - tasks 
that can be substituted by remote sensing. 
In order to reduce the costs of forest management planning, remote sensing techniques have be studied intensively for 
the last decades. The national forest inventory of Finland was the first of its kind utilizing satellite data (Tomppo, 1991). 
Satellite remote sensing has been shown to be an appropriate tool to assess and monitor large-area forest attributes with 
reasonable accuracy levels. Presently, optical satellite images are widely applied in national forest inventories. 
Concerning the stand level, no single remote sensing data acquisition method has been shown to be accurate and cost- 
effective enough for operational forest inventory, despite intensive research utilizing both airborne and space-borne 
remote sensing techniques (Poso et al., 1984; Pussinen, 1992; Tokola and Heikkild, 1997; Hyyppi et al., 2000a). Aerial 
photos have been widely accepted to assist in the process, but automatic derivation of stand parameters is constantly an 
aim of several research groups. Recently, promising results have been obtained by many authors using semi-automated 
forest inventory based on single tree crown delineation and aerial photos (Dralle and Rudemo, 1996; Gougeon, 1997; 
Brandtberg and Walter, 1998) or laser scanner (Hyyppä and Inkinen, 1999). 
The airborne laser scanner offers huge opportunities for rapid estimation of tree height, timber volume, and forest 
biomass over extensive forest areas. Previously, laser systems applied for forest studies were profiling sensors capable of 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 421