UTILIZATION OF MULTI TEMPORAL SAR DATA FOR FOREST MAPPING MODEL DEVELOPMENT

Abstract

Utilization  of  optical  satellite  data  in  tropical  region  was  limited to  free  cloud  cover. Therefore, Synthetic  Aperture  Radar  (SAR)  becomes  an  alternative  solution  for  forest  mapping  in Indonesia due to its capability to penetrate cloud. The objective of this research was to develop a forestmapping model based on multi temporal SAR data. Multi temporal ALOS PALSAR data for 2007 and 2008  were  used  for  forest  mapping,  and  one  year  mosaic  LANDSAT  data  in  2008  was  used  as references  data  to  obtain  training  sample  and  to  verify  the  final  forest  classification.  PALSAR processing was done using gamma naught conversion and Lee filtering. Samples were made in forest and  water  area, and  the  statistical  values  of the  each  object  were  calculated.  Some  thresholds  were determined  based  on  the  average  and  standard  deviation,  and  the  best  threshold  was  selected  to classify forest and water in 2008. It was assumed that forest could not change in 1-2 years period. The classification of forest, water, and the change were combined to produce final forest in 2008, and then it was visually verified with mosaic LANDSAT in 2008. The result showed that forest, water, and the change  could  be  well  classified  using  threshold  method.  The  forest  derived  from  PALSAR  was visually  consistent  with  forest  appearance  in  LANDSAT  and  forest  produced  from  INCAS.  It  has better performance than forest derived from INCAS for separating oil palm plantation from the forest.