Investigation on the capability of SPOT5 data for forest density mapping in Caspian forest

摘要

Forest canopy has a significant role in forest production, climate and ecosystem functions. It's also an essential quantitative characteristic for monitoring and sustainable management in Northern forests. In order to investigate the capability of SPOT-HRG (pan & xs)data for forest density mapping, the data of this sensor dated 2002 were analyzed. This study was carried out in Caspian forests of Iran. Qualitative and quantitative analysis of satellite data reveals no geometric and radiometric error. For accuracy assessment of classification results, ground truth covering 25% of the total area was prepared based on 1:40000 aerial photographs dated 2001.A total of 2520 circle sample plots with 1 ha area was selected and interpreted. The best spectral bands were defined based on the divergence criteria and sample areas. The supervised classification utilizing original and synthetic bands with maximum likelihood (ML)and minimum distance to mean (MD)classifier was performed. With 6 density classes (1-5%, 5-10%, 10-25%, 25-50% and >50%)and one non-forest class, the highest overall accuracy and kappa coefficient were 34% and 0.12 respectively. Signature separability, producer and user accuracies showed spectral similarity between density classes. After merging some classes, MLC exhibited the best results with 3 density classes (1-10%, 10-50%, >50%)and one non-forest class. The overall accuracy and kappa coefficient were estimated 74% and 0.33 respectively. To obtain better result, it is suggested to perform firstly, a forest non-forest classification and then to classify the forest into density classes. Using higher spectral resolution data are also suggested.