SPATIO-TEMPORAL ANALYSIS OF FOREST AREAS
Ivana Dabanovic
DGU- State Geodetic Administration, Riva 16, 51000 Rijeka, Croatia
- idabanovic@gmail.com
KEY WORDS: Landsat, Change detection, Classification, Land use, Multitemporal
ABSTRACT:
The main aim of the study was to determine the forest area change using Landsat ETM+ images in the observed time-span of 1990-
2000. After initial processing of images, optimal band combination was chosen based on data comparison with the digital orthophoto
of the area in the scale of 2000. Both supervised and unsupervised classification were being performed and compared to the same
aerial data. Change detection of forest areas is then followed by accuracy assessment. Software used was IDRISI Kilimanjaro.
1. INTRODUCTION
Forests are important natural and economic wealth and they
have big influence on human population on both global and
local level. Their dynamic change but also size itself makes
them an interesting topic for remote sensing application.
Growing population and specially urbanization of the popular
tourist areas like observed here could possibly reduce forests
span. Remote sensing is already widely used technique in
these assignments since results can be with a relatively high
accuracy, easy and fast to use from an expert but also with
cost much lower than any traditional method. Aim of the
study was to show the above stated and to produce data that
potentially can be used in land protection, land planning
purposes and also continued in future with up-to-date data. In
this work, specie of lat. Quercus robur L. was used for a time-
change analysis.
2. DATA USED
For receiving satellite images, I used free images of Landsat
from Institute for Advanced Computer Studies, University of
Maryland, USA site http://glcf.umiacs.umd.edu . This
institute makes preprocessing, radiometric or geometric
corrections within the images and I used data from June 6th
2000 and August 6th 1990.
Precisely, used 7-channel images are located on
ftp://ftp.glcf.umiacs.umd.edU/glcf/Landsat/WRS2/p 190/r029/
p 190r029_7x20000606.ETM-EarthSat-Orthorectified/ (190th
path and 29th row in the WRS system, dated June 6th 2000).
Image covers north of Adriatic sea and Croatia's west part. In
the preparation phase, following has been noticed: 1990.
images have poorer quality, small contrast and on 3rd and 4th
channel an obvious error of banding. Images from the year
2000. have area of clouds covering large part of terrain in
peninsula of Istra. For removing of banding, IDRISI's
Principal Components Analysis module was used and there
where 96% of data from first two components.
Picture 1: Overview
3. STUDY AREA
With the area of 409km 2 Krk and Cres share the title of the
largest islands in the Adriatic sea. Island Krk is on the far
north, and it has a well defined coastline and various
vegetation groups covering mainland. South of the island has
almost no vegetation, covered with up to 500m high bare
rock mountains.
This area has a moderate Mediterranean climate with some
above-average rainfall influenced by the proximity of Alps.
Aim of the study was to make a time-change analysis of the
forest areas on the island of Krk, represented mainly by
deciduous Quercus robur L and in smaller areas coniferous
pine forest.
Data used as reference data were digital orthophoto maps in Quercus robur L. is widely spread sort of oak on the island,
the scale of 1:2000 from the year of 2000 and positional and besides little parts of pine sorts in the south, we can say it
accuracy of+-10cm. is the only naturally grown tree. The density of the forest
varies, and the spatial resolution of the TM sensor does not
allow the detection of sparsely populated trees.
315
