AUTOMATIC FOREST AREA RECOGNITION USING GIS IMAGE ANALYSIS CAPABILITY 
Marco Ciolli, Paolo Zatelli 
Dipartimento di Ingegneria Civile ed Ambientale, Università degli Studi di Trento 
via Mesiano 77, 38100 Trento, tei. 0461.882678 - fax 0461.882672 
e-mail: Marco.Ciolli @ing.unitn.it - Paolo.Zatelli@ing.unitn.it 
KEY WORD: digitai image, classification, orthoimage, forest area recognition, GIS. 
ABSTRACT 
In the higher part of many Italian alpine regions the population is rapidly decreasing. This fact directly influences 
landscape and forest management criteria. In particular, large pastures and agricultured areas are being covered by 
trees or bushes, which change the landscape. This is a well known trend but it is very difficult to evaluate the extension 
and the evolution of these phenomena. A new approach to this problem has been developed at the Laboratorio Ambiente 
e Territorio, with which it is possible to determine changes of the forested areas over time. Series of aerial photographs 
taken in different years (1954, 1983, 1994) have been compared using automatic algorithms which exploit the GRASS 
GIS (Grass GIS is a public domain Geographic Information System) image analysis capability. The aerial photographs 
have been orthorectified to achieve images which can be directly superimposed to each other and to cartography. 
An image classification is carried out for each image using the following steps: preprocessing to identify homogeneous 
gray level areas; analysis of the gray level distribution in each area, filtering of the image to define boundaries of different 
kind of vegetation. A supervised homogenization is carried out to recover small errors, in particular small gaps inside 
wooden areas. 
The forest coverage evolution is finally computed by comparing the boundaries of the wooden areas in different years. 
A skilled operator is able to perform the whole operation in a very short time, especially compared to traditional 
techniques. This method has proved to be effective for the automatic determination of the widening of the forested areas. 
A precise knowledge of the forest coverage evolution is very precious to enhance forest and landscape management 
criteria. It can also be combined with demographic informal 
1. FOREST AREA RECOGNITION 
1.1 Introduction 
To carry out ecological landscape planning is necessary 
to dispose of many information about the development of 
the landscape through the years. The comprehension of 
the economic, social and cultural reasons which are tied 
to a landscape change gives to landscape planners and 
the technicians a correct point of view in land 
management. Sometimes it is very difficult to find 
information about landscape changes. Photogrammetry 
and Geographic Information System image analysis 
capability can be used to give a dramatic improvement to 
our knowledge of the land. 
Landscape and forest management criteria are influenced 
by the fact that in the higher part of many Italian alpine 
regions the population is rapidly decreasing. 
In particular, large pastures and agricultured areas are 
being covered by trees or bushes, which deeply change 
the landscape. This is a well known trend but it is very 
difficult to evaluate the extension and the evolution of the 
phenomena. This happens because foresters have no 
measures or direct samples which quantify the 
phenomena and which can be used to compare different 
situations at different times. 
A new approach to this problem has been developed at 
the Laboratorio Ambiente e Territorio of the Faculty of 
engineering of Trento making possible to determine 
changes of the forested areas over time. 
Ortho-photos are a very important historical document 
because they permit to read the land changes comparing 
different periods. The presence of aerial photographs 
taken in the fifties is particularly interesting because this 
was a period characterized by the most significant 
changes. 
to better explain landscape change. 
Series of aerial photos taken in different years (1954, 
1983, 1994) have been compared using automatic 
algorithms which exploit the GRASS GIS image analysis 
capability. 
1.2 Study area 
The study area has been chosen in Roncegno Mountain 
in the Roncegno Municipality, in the south east of the 
Trentino region (Italy). It is delimited by South by old 
Brenta River (mean height above sea level: 400 m), by 
East by Ceggio River, by North by Monte Gronlait and 
Monte Cola (2200 m above sea level) and by West by 
Larganza River. This region is rapidly changing both in 
socio-economic characters and in landscape. In this area, 
as well as in most of the Trentino region, a well known 
reforestation trend is present. The area has been chosen 
to verify that the available data are sufficient to carry out 
the quantification of the forest change. 
1.3 Tecniques 
Since the reforestation trend has never been quantified, 
this work is a first attempt to determine the extension of 
this phenomena by a quantitative analysis. 
Different solutions are possible. The first possibility is to 
carry out field surveys which present some problems: they 
are very expensive, they usually take a long time, they 
lack of unifomity and they are able to depict only the 
present situation. 
A second possibility relies on remote sensing which could 
be the best solution since data are collected in different 
spectral bands. Therefore a better forested area 
recognition can be exploited since the spectral signature 
of each specie can be identified.