Climate change and overall temperature increase results in changes in forest cover in high elevations. Due to the long life cycle of trees, these changes are very gradual and can be observed over long periods of time. In order to use remote sensing imagery for this purpose it needs to have very high spatial resolution and to have been acquired at least 50 years ago. At the moment, the only type of remote sensing imagery with these characteristics is historical black and white aerial photographs. This study used an aerial photograph from 1945 in order to map the forest cover at the Olympus National Park, at that date. An object-based classification (OBC) model was developed in order to classify forest and discriminate it from other types of vegetation. Due to the lack of near-infrared information, the model had to rely solely on the tone of the objects, as well as their geometric characteristics. The model functioned on three segmentation levels, using sub-/super-objects relationships and utilising vegetation density to discriminate forest and non-forest vegetation. The accuracy of the classification was assessed using 503 visually interpreted and randomly distributed points, resulting in a 92% overall accuracy. The model is using unbiased parameters that are important for differentiating between forest and non-forest vegetation and should be transferrable to other study areas of mountainous forests at high elevations.

Date Published: 12 August 2016
PDF: 7 pages
Proc. SPIE 9688, Fourth International Conference on Remote Sensing and Geoinformation of the Environment (RSCy2016), 96881H (12 August 2016); doi: 10.1117/12.2240738

Published in SPIE Proceedings Vol. 9688:
Fourth International Conference on Remote Sensing and Geoinformation of the Environment (RSCy2016)
Kyriacos Themistocleous; Diofantos G. Hadjimitsis; Silas Michaelides; Giorgos Papadavid, Editor(s)