Press release 9.8.2006
A technology has been developed at the Finnish Forest Research Institute for more reliable detection of thinnings using aerial photographs. By comparing aerial photographs taken in two different years 84% of thinned stands were classified correctly.
The aim of this Metla study was to establish whether bi-temporal aerial photography can improve the reliability of detecting the changes that have occurred in a forest. To do this, aerial photographs were taken from the same points at the same time of year and the same time of day. Additionally, resolution, radiometric correction, textual features, and stand attributes were also analyzed to determine their effects on the reliability of classification.
Two aerial photographs of the study area at Kauhajoki, in Southern Ostrobothnia were taken in the years 2001 and 2004. The aerial photographs were taken from almost the same target points at almost the same times. Thus, the unchanged stands looked almost identical in the images taken at the different times.
The major changes, such as regeneration cuttings, were found without error. In contrast, of the moderate changes, such as mild storm damage, seed tree cuttings, commercial thinnings, and tending of the seedling stand, 15-26 % were not detected. Of thinnings, 84% were correctly classified. On the other hand, of the unchanged stands, 14-25% was erroneously classified as moderately changed stands.
The results with respect to identifying the moderate changes were better in this study than in corresponding earlier studies. The biggest problem was the erroneous classification of unchanged stands as moderately changed stands. For the present, this prevents the technology from being implemented in practice.
The study is part of the "Continuous updating of forest resource information for forest management planning” project. Regional Forestry Centre Etelä-Pohjanmaa and the University of Joensuu have also participated in the project. During the study, which is still ongoing, the technology has been improved to increase the accuracy level: in the preliminary calculations 90% of thinnings have been detected and the share of erroneous classification as unchanged growing stocks as moderately changed ones has been reduced to almost 8%.
Publication: Hyvönen, P. & Anttila, P. 2006. Change detection in boreal forests using bi-temporal aerial photographs. Silva Fennica 40(2): 303–314.
The study summary in Finnish was published in Metsätieteen aikakauskirja 2/2006, p. 334-335. (pdf)