Forest density and orchard classification in Hyrcanian forests of Iran using Landsat 8 data

Abstract

Satellite-based remote sensing is of crucial importance to provide timely and continuous thematic maps for practical forestry tasks. There is currently no existing remote sensing-based, large-scale inventory of canopy cover classes (and also adjacent orchards) on the full range of Hyrcanian forests. We used the freely available and large-scale coverage of Landsat 8 imagery acquired in 2014 to classify three forest density classes as well as non-forest and orchards. The supervised classification and support vector machine classifier were selected based on a pre-classification of three representative pilot regions. Classified final maps were validated by means of a two-stage sampling and 1,852 field samples. The total areas of the dense, semi-dense, sparse forests and orchards were 45, 36, 19 and 1.9% of the total studied area, respectively. The overall accuracy and Kappa coefficient of classified maps were 94.8 and 90%, respectively. The methodology introduced to map forest cover in Hyrcanian forests is concluded to enable providing a high quality forest database for further research, planning and management.