Forest cover density mapping in sparse and semi dense forests using forest canopy density model (Case study: Marivan forests)

The main aim of this study was the evaluation of the Forest Canopy Density model (FCDm) for prediction of forest canopy density, using Landsat-7 ETM+. The study area was the eastern part of Marivan city that situated in Kurdistan province. A Landsat image was geo-referenced with sub pixel accuracy. First, all bands (1-5 of ETM+ imagery) except band 6 was normalized and then four main indices of FCD Model, including Advanced Vegetation Index, Bare soil Index, Shadow Index and thermal Index was calculated, and the forest canopy density map was derived finally. Forest's canopy densities according to 6, 4 and three classes were classified. To assess the accuracy of classified maps, a ground truth map using aerial photos with the scale 1:20000 was produced. The overall accuracy and kappa coefficient for classification 6 and four classes were obtained equal to 52%, 0.29 and 53%, 0.30, respectively. Spectral similarity between open density classes and irradiance of background soil in these classes reduced the accuracy as the result. Actually, in the dense forest, the result will be more accurate. According to the results, this method could be relatively desired for Zagro's forests.