In this study, performance of the Global 25m resolution ALOS-PALSAR mosaic and forest/non-forest map generated by Japan Space Exploration Agency's-JAXA was addressed. PALSAR imagery has dual polarimetric data (HH and HV) and these dataset are open and freely available. An additional band applying difference of two polarizations (HH-HV) was added as a third band. For the study area most populated city of Turkey Istanbul is selected due to its rapid and dense expanse. For the evaluation 2010 and 2015 mosaic PALSAR data was classified using k-Nearest Neighbors (k-NN) method as a practical image classification approach. Moreover, an accuracy analysis for the forest/non-forest maps was also investigated to expose its misleading interpretation. A change detection analyses was conducted to estimate the changes from 2010 to 2015. In addition to these, the classified data was compared with 30m GlobeLand30 global data product. Results indicated that high rate classification result as 86% and 89% overall was obtained by k-NN with the three bands of ALOS-PALSAR mosaic data of 2010 and 2015, respectively.