The most accurate method to estimate the SI is to measure the current height and age of a stand and then 
extrapolate the height to the reference age, following exactly the definition of the SI. However, to 
achieve acceptable precision, the stand has to satisfy several requirements. It must be even-aged, 
undamaged, have one dominating species etc. Unfortunately, relatively few stands fulfill these 
requirements. 
Therefore, an alternative method was developed that can be used in all sorts of stands. The site is 
described by a number of variables supposed to indicate the tree growth potential. Of these, the most 
significant are: latitude, altitude, vegetation type, soil texture, and ground moisture. Regression 
functions then convert the variables to a site index. This method is the most commonly used and thereby 
the norm for management planning, land valuation, and law enforcement in Swedish forestry. It is 
referenced later in this paper as the "existing SCS". 
Digital Site Classification, Project Background and Goals. 
The project in question started in late 1989. The overall goal is to develop a new digital method to 
assess the site quality by land information analysis. The work focuses on the following two 
disadvantages in the existing SCS: 
- To estimate the SI, it is necessary to stand at the specific site and make manual measurements. No 
indoor method exists today. This is ineffective since experienced field personnel are rare and 
expensive. 
- The precision is rather low with a mean deviation as high as 20% on the estimated volume growth. The 
two main reasons are: 1. The site variables have been simplified in order to fit the manual 
measurements in the field. The variables therefore do not accurately describe the growth conditions. 
2. The manual measurements include subjective aspects which can lead to low precision. 
At the Remote Sensing Laboratory of the Faculty of Forestry, research mainly concerns how to extract 
forest information out of satellite imagery (HAGNER 1989). However, the question often arises as to how 
the enormous amount of land information available in Sweden, such as Digital Elevation Models (DEMs), 
soil maps, geological maps, cadastral base maps etc. could be utilized for forest inventory purposes. 
The basic assumption is that it probably would be possible to digitally classify the site index for each 
pixel (e.g. 10x10m) over a large land area using available map information. The resulting raster image 
would be an extremely useful complement to satellite images when estimating forest characteristics. In 
fact, the image could be used as just another band along with the traditional spectral ones. 
Thus, the goals for the project can be summarized as: 
- to improve the precision in site quality classification by incorporating more complex and more 
relevant site variables, not possible to measure in the field, but possible to extract from DEMs and 
other digital land information. 
- to rationalize site classification by decreasing the need for field checks. 
- to produce complementary image information on the site quality, to be used along with other digital 
(satellite) images for classification purposes. 
Estimating site variables for the existing SCS. 
This paper reports on a pilot study which evaluated the possibility to determine site variables for the 
existing SCS from digital map materials. The goals were: 
871